Add @top to some misc/*.texi files.

[emacs.git] / etc / PROBLEMS

Known Problems with GNU Emacs

Copyright (C) 1987-1989, 1993-1999, 2001-2011
  Free Software Foundation, Inc.
See the end of the file for license conditions.


This file describes various problems that have been encountered
in compiling, installing and running GNU Emacs.  Try doing C-c C-t
and browsing through the outline headers.  (See C-h m for help on
Outline mode.)  Information about systems that are no longer supported,
and old Emacs releases, has been removed.  Consult older versions of
this file if you are interested in that information.

* Mule-UCS doesn't work in Emacs 23.

It's completely redundant now, as far as we know.

* Emacs startup failures

** Emacs fails to start, complaining about missing fonts.

A typical error message might be something like

  No fonts match `-*-fixed-medium-r-*--6-*-*-*-*-*-iso8859-1'

This happens because some X resource specifies a bad font family for
Emacs to use.  The possible places where this specification might be
are:

  - in your ~/.Xdefaults file

  - client-side X resource file, such as  ~/Emacs or
    /usr/X11R6/lib/app-defaults/Emacs or
    /usr/X11R6/lib/X11/app-defaults/Emacs

One of these files might have bad or malformed specification of a
fontset that Emacs should use.  To fix the problem, you need to find
the problematic line(s) and correct them.

** Emacs aborts while starting up, only when run without X.

This problem often results from compiling Emacs with GCC when GCC was
installed incorrectly.  The usual error in installing GCC is to
specify --includedir=/usr/include.  Installation of GCC makes
corrected copies of the system header files.  GCC is supposed to use
the corrected copies in preference to the original system headers.
Specifying --includedir=/usr/include causes the original system header
files to be used.  On some systems, the definition of ioctl in the
original system header files is invalid for ANSI C and causes Emacs
not to work.

The fix is to reinstall GCC, and this time do not specify --includedir
when you configure it.  Then recompile Emacs.  Specifying --includedir
is appropriate only in very special cases and it should *never* be the
same directory where system header files are kept.

** Emacs does not start, complaining that it cannot open termcap database file.

If your system uses Terminfo rather than termcap (most modern
systems do), this could happen if the proper version of
ncurses is not visible to the Emacs configure script (i.e. it
cannot be found along the usual path the linker looks for
libraries). It can happen because your version of ncurses is
obsolete, or is available only in form of binaries.

The solution is to install an up-to-date version of ncurses in
the developer's form (header files, static libraries and
symbolic links); in some GNU/Linux distributions (e.g. Debian)
it constitutes a separate package.

** Emacs 20 and later fails to load Lisp files at startup.

The typical error message might be like this:

  "Cannot open load file: fontset"

This could happen if you compress the file lisp/subdirs.el.  That file
tells Emacs what are the directories where it should look for Lisp
files.  Emacs cannot work with subdirs.el compressed, since the
Auto-compress mode it needs for this will not be loaded until later,
when your .emacs file is processed.  (The package `fontset.el' is
required to set up fonts used to display text on window systems, and
it's loaded very early in the startup procedure.)

Similarly, any other .el file for which there's no corresponding .elc
file could fail to load if it is compressed.

The solution is to uncompress all .el files that don't have a .elc file.

Another possible reason for such failures is stale *.elc files
lurking somewhere on your load-path -- see the next section.

** Emacs prints an error at startup after upgrading from an earlier version.

An example of such an error is:

  x-complement-fontset-spec: "Wrong type argument: stringp, nil"

This can be another symptom of stale *.elc files in your load-path.
The following command will print any duplicate Lisp files that are
present in load-path:

    emacs -batch -f list-load-path-shadows

If this command prints any file names, some of these files are stale,
and should be deleted or their directories removed from your
load-path.

** With X11R6.4, public-patch-3, Emacs crashes at startup.

Reportedly this patch in X fixes the problem.

    --- xc/lib/X11/imInt.c~     Wed Jun 30 13:31:56 1999
    +++ xc/lib/X11/imInt.c      Thu Jul  1 15:10:27 1999
    @@ -1,4 +1,4 @@
    -/* $TOG: imInt.c /main/5 1998/05/30 21:11:16 kaleb $ */
    +/* $TOG: imInt.c /main/5 1998/05/30 21:11:16 kaleb $  */
     /******************************************************************

                Copyright 1992, 1993, 1994 by FUJITSU LIMITED
    @@ -166,8 +166,8 @@
     _XimMakeImName(lcd)
         XLCd      lcd;
     {
    -    char* begin;
    -    char* end;
    +    char* begin = NULL;
    +    char* end = NULL;
         char* ret;
         int    i = 0;
         char* ximmodifier = XIMMODIFIER;
    @@ -182,7 +182,11 @@
         }
         ret = Xmalloc(end - begin + 2);
         if (ret != NULL) {
    -           (void)strncpy(ret, begin, end - begin + 1);
    +   if (begin != NULL) {
    +             (void)strncpy(ret, begin, end - begin + 1);
    +        } else {
    +     ret[0] = '\0';
    +   }
            ret[end - begin + 1] = '\0';
         }
         return ret;

** Emacs crashes on startup after a glibc upgrade.

This is caused by a binary incompatible change to the malloc
implementation in glibc 2.5.90-22.  As a result, Emacs binaries built
using prior versions of glibc crash when run under 2.5.90-22.

This problem was first seen in pre-release versions of Fedora 7, and
may be fixed in the final Fedora 7 release.  To stop the crash from
happening, first try upgrading to the newest version of glibc; if this
does not work, rebuild Emacs with the same version of glibc that you
will run it under.  For details, see

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=239344

* Crash bugs

** Emacs crashes when running in a terminal, if compiled with GCC 4.5.0
This version of GCC is buggy: see

  http://debbugs.gnu.org/cgi/bugreport.cgi?bug=6031
  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43904

You can work around this error in gcc-4.5 by omitting sibling call
optimization.  To do this, configure Emacs with

 CFLAGS="-g -O2 -fno-optimize-sibling-calls" ./configure

** Emacs crashes in x-popup-dialog.

This can happen if the dialog widget cannot find the font it wants to
use.  You can work around the problem by specifying another font with
an X resource--for example, `Emacs.dialog*.font: 9x15' (or any font that
happens to exist on your X server).

** Emacs crashes when you use Bibtex mode.

This happens if your system puts a small limit on stack size.  You can
prevent the problem by using a suitable shell command (often `ulimit')
to raise the stack size limit before you run Emacs.

Patches to raise the stack size limit automatically in `main'
(src/emacs.c) on various systems would be greatly appreciated.

** Error message `Symbol's value as variable is void: x', followed by
a segmentation fault and core dump.

This has been tracked to a bug in tar!  People report that tar erroneously
added a line like this at the beginning of files of Lisp code:

   x FILENAME, N bytes, B tape blocks

If your tar has this problem, install GNU tar--if you can manage to
untar it :-).

** Crashes when displaying GIF images in Emacs built with version
libungif-4.1.0 are resolved by using version libungif-4.1.0b1.
Configure checks for the correct version, but this problem could occur
if a binary built against a shared libungif is run on a system with an
older version.

** Emacs aborts inside the function `tparam1'.

This can happen if Emacs was built without terminfo support, but the
terminal's capabilities use format that is only supported by terminfo.
If your system has ncurses installed, this might happen if your
version of ncurses is broken; upgrading to a newer version of ncurses
and reconfiguring and rebuilding Emacs should solve this.

All modern systems support terminfo, so even if ncurses is not the
problem, you should look for a way to configure Emacs so that it uses
terminfo when built.

** Emacs crashes when using some version of the Exceed X server.

Upgrading to a newer version of Exceed has been reported to prevent
these crashes.  You should consider switching to a free X server, such
as Xming or Cygwin/X.

** Emacs crashes with SIGSEGV in XtInitializeWidgetClass.

It crashes on X, but runs fine when called with option "-nw".

This has been observed when Emacs is linked with GNU ld but without passing
the -z nocombreloc flag.  Emacs normally knows to pass the -z nocombreloc
flag when needed, so if you come across a situation where the flag is
necessary but missing, please report it via M-x report-emacs-bug.

On platforms such as Solaris, you can also work around this problem by
configuring your compiler to use the native linker instead of GNU ld.

** When Emacs is compiled with Gtk+, closing a display kills Emacs.

There is a long-standing bug in GTK that prevents it from recovering
from disconnects: http://bugzilla.gnome.org/show_bug.cgi?id=85715.

Thus, for instance, when Emacs is run as a server on a text terminal,
and an X frame is created, and the X server for that frame crashes or
exits unexpectedly, Emacs must exit to prevent a GTK error that would
result in an endless loop.

If you need Emacs to be able to recover from closing displays, compile
it with the Lucid toolkit instead of GTK.

* General runtime problems

** Lisp problems

*** Changes made to .el files do not take effect.

You may have forgotten to recompile them into .elc files.
Then the old .elc files will be loaded, and your changes
will not be seen.  To fix this, do M-x byte-recompile-directory
and specify the directory that contains the Lisp files.

Emacs should print a warning when loading a .elc file which is older
than the corresponding .el file.

*** Watch out for .emacs files and EMACSLOADPATH environment vars.

These control the actions of Emacs.
~/.emacs is your Emacs init file.
EMACSLOADPATH overrides which directories the function "load" will search.

If you observe strange problems, check for these and get rid
of them, then try again.

*** Using epop3.el package causes Emacs to signal an error.

The error message might be something like this:

  "Lisp nesting exceeds max-lisp-eval-depth"

This happens because epop3 redefines the function gethash, which is a
built-in primitive beginning with Emacs 21.1.  We don't have a patch
for epop3 that fixes this, but perhaps a newer version of epop3
corrects that.

*** Buffers from `with-output-to-temp-buffer' get set up in Help mode.

Changes in Emacs 20.4 to the hooks used by that function cause
problems for some packages, specifically BBDB.  See the function's
documentation for the hooks involved.  BBDB 2.00.06 fixes the problem.

*** The Hyperbole package causes *Help* buffers not to be displayed in
Help mode due to setting `temp-buffer-show-hook' rather than using
`add-hook'.  Using `(add-hook 'temp-buffer-show-hook
'help-mode-maybe)' after loading Hyperbole should fix this.

** Keyboard problems

*** "Compose Character" key does strange things when used as a Meta key.

If you define one key to serve as both Meta and Compose Character, you
will get strange results.  In previous Emacs versions, this "worked"
in that the key acted as Meta--that's because the older Emacs versions
did not try to support Compose Character.  Now Emacs tries to do
character composition in the standard X way.  This means that you
must pick one meaning or the other for any given key.

You can use both functions (Meta, and Compose Character) if you assign
them to two different keys.

*** C-z just refreshes the screen instead of suspending Emacs.

You are probably using a shell that doesn't support job control, even
though the system itself is capable of it.  Either use a different shell,
or set the variable `cannot-suspend' to a non-nil value.

*** With M-x enable-flow-control, you need to type C-\ twice
to do incremental search--a single C-\ gets no response.

This has been traced to communicating with your machine via kermit,
with C-\ as the kermit escape character.  One solution is to use
another escape character in kermit.  One user did

   set escape-character 17

in his .kermrc file, to make C-q the kermit escape character.

** Mailers and other helper programs

*** movemail compiled with POP support can't connect to the POP server.

Make sure that the `pop' entry in /etc/services, or in the services
NIS map if your machine uses NIS, has the same port number as the
entry on the POP server.  A common error is for the POP server to be
listening on port 110, the assigned port for the POP3 protocol, while
the client is trying to connect on port 109, the assigned port for the
old POP protocol.

*** RMAIL gets error getting new mail.

RMAIL gets new mail from /usr/spool/mail/$USER using a program
called `movemail'.  This program interlocks with /bin/mail using
the protocol defined by /bin/mail.

There are two different protocols in general use.  One of them uses
the `flock' system call.  The other involves creating a lock file;
`movemail' must be able to write in /usr/spool/mail in order to do
this.  You control which one is used by defining, or not defining,
the macro MAIL_USE_FLOCK in config.h or the m- or s- file it includes.
IF YOU DON'T USE THE FORM OF INTERLOCKING THAT IS NORMAL ON YOUR
SYSTEM, YOU CAN LOSE MAIL!

If your system uses the lock file protocol, and fascist restrictions
prevent ordinary users from writing the lock files in /usr/spool/mail,
you may need to make `movemail' setgid to a suitable group such as
`mail'.  To do this,  use the following commands (as root) after doing the
make install.

        chgrp mail movemail
        chmod 2755 movemail

Installation normally copies movemail from the build directory to an
installation directory which is usually under /usr/local/lib.  The
installed copy of movemail is usually in the directory
/usr/local/lib/emacs/VERSION/TARGET.  You must change the group and
mode of the installed copy; changing the group and mode of the build
directory copy is ineffective.

*** rcs2log gives you the awk error message "too many fields".

This is due to an arbitrary limit in certain versions of awk.
The solution is to use gawk (GNU awk).

** Problems with hostname resolution

*** Emacs fails to understand most Internet host names, even though
the names work properly with other programs on the same system.
*** Emacs won't work with X-windows if the value of DISPLAY is HOSTNAME:0.
*** Gnus can't make contact with the specified host for nntp.

This typically happens on Suns and other systems that use shared
libraries.  The cause is that the site has installed a version of the
shared library which uses a name server--but has not installed a
similar version of the unshared library which Emacs uses.

The result is that most programs, using the shared library, work with
the nameserver, but Emacs does not.

The fix is to install an unshared library that corresponds to what you
installed in the shared library, and then relink Emacs.

If you have already installed the name resolver in the file libresolv.a,
then you need to compile Emacs to use that library.  The easiest way to
do this is to add to config.h a definition of LIBS_SYSTEM, LIBS_MACHINE
or LIB_STANDARD which uses -lresolv.  Watch out!  If you redefine a macro
that is already in use in your configuration to supply some other libraries,
be careful not to lose the others.

Thus, you could start by adding this to config.h:

#define LIBS_SYSTEM -lresolv

Then if this gives you an error for redefining a macro, and you see that
the s- file defines LIBS_SYSTEM as -lfoo -lbar, you could change config.h
again to say this:

#define LIBS_SYSTEM -lresolv -lfoo -lbar

*** Emacs does not know your host's fully-qualified domain name.

For example, (system-name) returns some variation on
"localhost.localdomain", rather the name you were expecting.

You need to configure your machine with a fully qualified domain name,
(i.e. a name with at least one ".") either in /etc/hosts,
/etc/hostname, the NIS, or wherever your system calls for specifying this.

If you cannot fix the configuration, you can set the Lisp variable
mail-host-address to the value you want.

** NFS and RFS

*** Emacs says it has saved a file, but the file does not actually
appear on disk.

This can happen on certain systems when you are using NFS, if the
remote disk is full.  It is due to a bug in NFS (or certain NFS
implementations), and there is apparently nothing Emacs can do to
detect the problem.  Emacs checks the failure codes of all the system
calls involved in writing a file, including `close'; but in the case
where the problem occurs, none of those system calls fails.

*** Editing files through RFS gives spurious "file has changed" warnings.
It is possible that a change in Emacs 18.37 gets around this problem,
but in case not, here is a description of how to fix the RFS bug that
causes it.

    There was a serious pair of bugs in the handling of the fsync() system
    call in the RFS server.

    The first is that the fsync() call is handled as another name for the
    close() system call (!!).  It appears that fsync() is not used by very
    many programs; Emacs version 18 does an fsync() before closing files
    to make sure that the bits are on the disk.

    This is fixed by the enclosed patch to the RFS server.

    The second, more serious problem, is that fsync() is treated as a
    non-blocking system call (i.e., it's implemented as a message that
    gets sent to the remote system without waiting for a reply).  Fsync is
    a useful tool for building atomic file transactions.  Implementing it
    as a non-blocking RPC call (when the local call blocks until the sync
    is done) is a bad idea; unfortunately, changing it will break the RFS
    protocol.  No fix was supplied for this problem.

    (as always, your line numbers may vary)

    % rcsdiff -c -r1.2 serversyscall.c
    RCS file: RCS/serversyscall.c,v
    retrieving revision 1.2
    diff -c -r1.2 serversyscall.c
    *** /tmp/,RCSt1003677   Wed Jan 28 15:15:02 1987
    --- serversyscall.c     Wed Jan 28 15:14:48 1987
    ***************
    *** 163,169 ****
            /*
             * No return sent for close or fsync!
             */
    !       if (syscall == RSYS_close || syscall == RSYS_fsync)
                    proc->p_returnval = deallocate_fd(proc, msg->m_args[0]);
            else
            {
    --- 166,172 ----
            /*
             * No return sent for close or fsync!
             */
    !       if (syscall == RSYS_close)
                    proc->p_returnval = deallocate_fd(proc, msg->m_args[0]);
            else
            {

** PSGML conflicts with sgml-mode.

PSGML package uses the same names of some variables (like keymap)
as built-in sgml-mode.el because it was created as a replacement
of that package.  The conflict will be shown if you load
sgml-mode.el before psgml.el.  E.g. this could happen if you edit
HTML page and then start to work with SGML or XML file.  html-mode
(from sgml-mode.el) is used for HTML file and loading of psgml.el
(for sgml-mode or xml-mode) will cause an error.

** PCL-CVS

*** Lines are not updated or new lines are added in the buffer upon commit.

When committing files located higher in the hierarchy than the examined
directory, some versions of the CVS program return an ambiguous message
from which PCL-CVS cannot extract the full location of the committed
files.  As a result, the corresponding lines in the PCL-CVS buffer are
not updated with the new revision of these files, and new lines are
added to the top-level directory.

This can happen with CVS versions 1.12.8 and 1.12.9.  Upgrade to CVS
1.12.10 or newer to fix this problem.

** Miscellaneous problems

*** Emacs uses 100% of CPU time

This is a known problem with some versions of the Semantic package.
The solution is to upgrade Semantic to version 2.0pre4 (distributed
with CEDET 1.0pre4) or later.

*** Self-documentation messages are garbled.

This means that the file `etc/DOC-...' doesn't properly correspond
with the Emacs executable.  Redumping Emacs and then installing the
corresponding pair of files should fix the problem.

*** Programs running under terminal emulator do not recognize `emacs'
terminal type.

The cause of this is a shell startup file that sets the TERMCAP
environment variable.  The terminal emulator uses that variable to
provide the information on the special terminal type that Emacs emulates.

Rewrite your shell startup file so that it does not change TERMCAP
in such a case.  You could use the following conditional which sets
it only if it is undefined.

    if ( ! ${?TERMCAP} ) setenv TERMCAP ~/my-termcap-file

Or you could set TERMCAP only when you set TERM--which should not
happen in a non-login shell.

*** In Shell mode, you get a ^M at the end of every line.

This happens to people who use tcsh, because it is trying to be too
smart.  It sees that the Shell uses terminal type `unknown' and turns
on the flag to output ^M at the end of each line.  You can fix the
problem by adding this to your .cshrc file:

    if ($?EMACS) then
        if ("$EMACS" =~ /*) then
            unset edit
            stty  -icrnl -onlcr -echo susp ^Z
        endif
    endif

*** Emacs startup on GNU/Linux systems (and possibly other systems) is slow.

This can happen if the system is misconfigured and Emacs can't get the
full qualified domain name, FQDN.  You should have your FQDN in the
/etc/hosts file, something like this:

127.0.0.1       localhost
129.187.137.82  nuc04.t30.physik.tu-muenchen.de nuc04

The way to set this up may vary on non-GNU systems.

*** Attempting to visit remote files via ange-ftp fails.

If the error message is "ange-ftp-file-modtime: Specified time is not
representable", then this could happen when `lukemftp' is used as the
ftp client.  This was reported to happen on Debian GNU/Linux, kernel
version 2.4.3, with `lukemftp' 1.5-5, but might happen on other
systems as well.  To avoid this problem, switch to using the standard
ftp client.  On a Debian system, type

  update-alternatives --config ftp

and then choose /usr/bin/netkit-ftp.

*** JPEG images aren't displayed.

This has been reported when Emacs is built with jpeg-6a library.
Upgrading to jpeg-6b solves the problem.  Configure checks for the
correct version, but this problem could occur if a binary built
against a shared libjpeg is run on a system with an older version.

*** Dired is very slow.

This could happen if invocation of the `df' program takes a long
time.  Possible reasons for this include:

  - ClearCase mounted filesystems (VOBs) that sometimes make `df'
    response time extremely slow (dozens of seconds);

  - slow automounters on some old versions of Unix;

  - slow operation of some versions of `df'.

To work around the problem, you could either (a) set the variable
`directory-free-space-program' to nil, and thus prevent Emacs from
invoking `df'; (b) use `df' from the GNU Fileutils package; or
(c) use CVS, which is Free Software, instead of ClearCase.

*** ps-print commands fail to find prologue files ps-prin*.ps.

This can happen if you use an old version of X-Symbol package: it
defines compatibility functions which trick ps-print into thinking it
runs in XEmacs, and look for the prologue files in a wrong directory.

The solution is to upgrade X-Symbol to a later version.

*** On systems with shared libraries you might encounter run-time errors
from the dynamic linker telling you that it is unable to find some
shared libraries, for instance those for Xaw3d or image support.
These errors mean Emacs has been linked with a library whose shared
library is not in the default search path of the dynamic linker.

Similar problems could prevent Emacs from building, since the build
process invokes Emacs several times.

On many systems, it is possible to set LD_LIBRARY_PATH in your
environment to specify additional directories where shared libraries
can be found.

Other systems allow to set LD_RUN_PATH in a similar way, but before
Emacs is linked.  With LD_RUN_PATH set, the linker will include a
specified run-time search path in the executable.

On some systems, Emacs can crash due to problems with dynamic
linking.  Specifically, on SGI Irix 6.5, crashes were reported with
backtraces like this:

  (dbx) where
   0 strcmp(0xf49239d, 0x4031184, 0x40302b4, 0x12, 0xf0000000, 0xf4923aa, 0x0, 0x492ddb2) ["/xlv22/ficus-jan23/work/irix/lib/libc/libc_n32_M3_ns/strings/strcmp.s":35, 0xfb7e480]
   1 general_find_symbol(0xf49239d, 0x0, 0x0, 0x0, 0xf0000000, 0xf4923aa, 0x0, 0x492ddb2)
 ["/comp2/mtibuild/v73/workarea/v7.3/rld/rld.c":2140, 0xfb65a98]
   2 resolve_symbol(0xf49239d, 0x4031184, 0x0, 0xfbdd438, 0x0, 0xf4923aa, 0x0, 0x492ddb2)
 ["/comp2/mtibuild/v73/workarea/v7.3/rld/rld.c":1947, 0xfb657e4]
   3 lazy_text_resolve(0xd18, 0x1a3, 0x40302b4, 0x12, 0xf0000000, 0xf4923aa, 0x0, 0x492ddb2)
 ["/comp2/mtibuild/v73/workarea/v7.3/rld/rld.c":997, 0xfb64d44]
   4 _rld_text_resolve(0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0)
 ["/comp2/mtibuild/v73/workarea/v7.3/rld/rld_bridge.s":175, 0xfb6032c]

(`rld' is the dynamic linker.)  We don't know yet why this
happens, but setting the environment variable LD_BIND_NOW to 1 (which
forces the dynamic linker to bind all shared objects early on) seems
to work around the problem.

Please refer to the documentation of your dynamic linker for details.

*** You request inverse video, and the first Emacs frame is in inverse
video, but later frames are not in inverse video.

This can happen if you have an old version of the custom library in
your search path for Lisp packages.  Use M-x list-load-path-shadows to
check whether this is true.  If it is, delete the old custom library.

*** When you run Ispell from Emacs, it reports a "misalignment" error.

This can happen if you compiled the Ispell program to use ASCII
characters only and then try to use it from Emacs with non-ASCII
characters, like Latin-1.  The solution is to recompile Ispell with
support for 8-bit characters.

To see whether your Ispell program supports 8-bit characters, type
this at your shell's prompt:

     ispell -vv

and look in the output for the string "NO8BIT".  If Ispell says
"!NO8BIT (8BIT)", your speller supports 8-bit characters; otherwise it
does not.

To rebuild Ispell with 8-bit character support, edit the local.h file
in the Ispell distribution and make sure it does _not_ define NO8BIT.
Then rebuild the speller.

Another possible cause for "misalignment" error messages is that the
version of Ispell installed on your machine is old.  Upgrade.

Yet another possibility is that you are trying to spell-check a word
in a language that doesn't fit the dictionary you choose for use by
Ispell.  (Ispell can only spell-check one language at a time, because
it uses a single dictionary.)  Make sure that the text you are
spelling and the dictionary used by Ispell conform to each other.

If your spell-checking program is Aspell, it has been reported that if
you have a personal configuration file (normally ~/.aspell.conf), it
can cause this error.  Remove that file, execute `ispell-kill-ispell'
in Emacs, and then try spell-checking again.

* Runtime problems related to font handling

** Characters are displayed as empty boxes or with wrong font under X.

*** This can occur when two different versions of FontConfig are used.
For example, XFree86 4.3.0 has one version and Gnome usually comes
with a newer version.  Emacs compiled with Gtk+ will then use the
newer version.  In most cases the problem can be temporarily fixed by
stopping the application that has the error (it can be Emacs or any
other application), removing ~/.fonts.cache-1, and then start the
application again.  If removing ~/.fonts.cache-1 and restarting
doesn't help, the application with problem must be recompiled with the
same version of FontConfig as the rest of the system uses.  For KDE,
it is sufficient to recompile Qt.

*** Some fonts have a missing glyph and no default character.  This is
known to occur for character number 160 (no-break space) in some
fonts, such as Lucida but Emacs sets the display table for the unibyte
and Latin-1 version of this character to display a space.

*** Some of the fonts called for in your fontset may not exist on your
X server.

Each X11 font covers just a fraction of the characters that Emacs
supports.  To display the whole range of Emacs characters requires
many different fonts, collected into a fontset.  You can remedy the
problem by installing additional fonts.

The intlfonts distribution includes a full spectrum of fonts that can
display all the characters Emacs supports.  The etl-unicode collection
of fonts (available from <URL:ftp://ftp.x.org/contrib/fonts/> and
<URL:ftp://ftp.xfree86.org/pub/mirror/X.Org/contrib/fonts/>) includes
fonts that can display many Unicode characters; they can also be used
by ps-print and ps-mule to print Unicode characters.

** Under X11, some characters appear improperly aligned in their lines.

You may have bad X11 fonts; try installing the intlfonts distribution
or the etl-unicode collection (see above).

** Under X, an unexpected monospace font is used as the default font.

When compiled with XFT, Emacs tries to use a default font named
"monospace".  This is a "virtual font", which the operating system
(Fontconfig) redirects to a suitable font such as DejaVu Sans Mono.
On some systems, there exists a font that is actually named Monospace,
which takes over the virtual font.  This is considered an operating
system bug; see

http://lists.gnu.org/archive/html/emacs-devel/2008-10/msg00696.html

If you encounter this problem, set the default font to a specific font
in your .Xresources or initialization file.  For instance, you can put
the following in your .Xresources:

Emacs.font: DejaVu Sans Mono 12

** Certain fonts make each line take one pixel more than it should.

This is because these fonts contain characters a little taller than
the font's nominal height.  Emacs needs to make sure that lines do not
overlap.

** Loading fonts is very slow.

You might be getting scalable fonts instead of precomputed bitmaps.
Known scalable font directories are "Type1" and "Speedo".  A font
directory contains scalable fonts if it contains the file
"fonts.scale".

If this is so, re-order your X windows font path to put the scalable
font directories last.  See the documentation of `xset' for details.

With some X servers, it may be necessary to take the scalable font
directories out of your path entirely, at least for Emacs 19.26.
Changes in the future may make this unnecessary.

** Font Lock displays portions of the buffer in incorrect faces.

By far the most frequent cause of this is a parenthesis `(' or a brace
`{' in column zero.  Font Lock assumes that such a paren is outside of
any comment or string.  This is of course not true in general, but the
vast majority of well-formatted program source files don't have such
parens, and therefore this assumption is used to allow optimizations
in Font Lock's syntactical analysis.  These optimizations avoid some
pathological cases where jit-lock, the Just-in-Time fontification
introduced with Emacs 21.1, could significantly slow down scrolling
through the buffer, especially scrolling backwards, and also jumping
to the end of a very large buffer.

Beginning with version 22.1, a parenthesis or a brace in column zero
is highlighted in bold-red face if it is inside a string or a comment,
to indicate that it could interfere with Font Lock (and also with
indentation) and should be moved or escaped with a backslash.

If you don't use large buffers, or have a very fast machine which
makes the delays insignificant, you can avoid the incorrect
fontification by setting the variable
`font-lock-beginning-of-syntax-function' to a nil value.  (This must
be done _after_ turning on Font Lock.)

Another alternative is to avoid a paren in column zero.  For example,
in a Lisp string you could precede the paren with a backslash.

** With certain fonts, when the cursor appears on a character, the
character doesn't appear--you get a solid box instead.

One user on a Linux-based GNU system reported that this problem went
away with installation of a new X server.  The failing server was
XFree86 3.1.1.  XFree86 3.1.2 works.

** Emacs pauses for several seconds when changing the default font.

This has been reported for fvwm 2.2.5 and the window manager of KDE
2.1.  The reason for the pause is Xt waiting for a ConfigureNotify
event from the window manager, which the window manager doesn't send.
Xt stops waiting after a default timeout of usually 5 seconds.

A workaround for this is to add something like

emacs.waitForWM: false

to your X resources.  Alternatively, add `(wait-for-wm . nil)' to a
frame's parameter list, like this:

   (modify-frame-parameters nil '((wait-for-wm . nil)))

(this should go into your `.emacs' file).

** Underlines appear at the wrong position.

This is caused by fonts having a wrong UNDERLINE_POSITION property.
Examples are the font 7x13 on XFree prior to version 4.1, or the jmk
neep font from the Debian xfonts-jmk package prior to version 3.0.17.
To circumvent this problem, set x-use-underline-position-properties
to nil in your `.emacs'.

To see what is the value of UNDERLINE_POSITION defined by the font,
type `xlsfonts -lll FONT' and look at the font's UNDERLINE_POSITION property.

** When using Exceed, fonts sometimes appear too tall.

When the display is set to an Exceed X-server and fonts are specified
(either explicitly with the -fn option or implicitly with X resources)
then the fonts may appear "too tall".  The actual character sizes are
correct but there is too much vertical spacing between rows,  which
gives the appearance of "double spacing".

To prevent this, turn off the Exceed's "automatic font substitution"
feature (in the font part of the configuration window).

** Subscript/superscript text in TeX is hard to read.

If `tex-fontify-script' is non-nil, tex-mode displays
subscript/superscript text in the faces subscript/superscript, which
are smaller than the normal font and lowered/raised.  With some fonts,
nested superscripts (say) can be hard to read.  Switching to a
different font, or changing your antialiasing setting (on an LCD
screen), can both make the problem disappear.  Alternatively, customize
the following variables: tex-font-script-display (how much to
lower/raise); tex-suscript-height-ratio (how much smaller than
normal); tex-suscript-height-minimum (minimum height).

* Internationalization problems

** M-{ does not work on a Spanish PC keyboard.

Many Spanish keyboards seem to ignore that combination.  Emacs can't
do anything about it.

** International characters aren't displayed under X.

*** Missing X fonts

XFree86 4 contains many fonts in iso10646-1 encoding which have
minimal character repertoires (whereas the encoding part of the font
name is meant to be a reasonable indication of the repertoire
according to the XLFD spec).  Emacs may choose one of these to display
characters from the mule-unicode charsets and then typically won't be
able to find the glyphs to display many characters.  (Check with C-u
C-x = .)  To avoid this, you may need to use a fontset which sets the
font for the mule-unicode sets explicitly.  E.g. to use GNU unifont,
include in the fontset spec:

mule-unicode-2500-33ff:-gnu-unifont-*-iso10646-1,\
mule-unicode-e000-ffff:-gnu-unifont-*-iso10646-1,\
mule-unicode-0100-24ff:-gnu-unifont-*-iso10646-1

** The UTF-8/16/7 coding systems don't encode CJK (Far Eastern) characters.

Emacs directly supports the Unicode BMP whose code points are in the
ranges 0000-33ff and e000-ffff, and indirectly supports the parts of
CJK characters belonging to these legacy charsets:

    GB2312, Big5, JISX0208, JISX0212, JISX0213-1, JISX0213-2, KSC5601

The latter support is done in Utf-Translate-Cjk mode (turned on by
default).   Which Unicode CJK characters are decoded into which Emacs
charset is decided by the current language environment.  For instance,
in Chinese-GB, most of them are decoded into chinese-gb2312.

If you read UTF-8 data with code points outside these ranges, the
characters appear in the buffer as raw bytes of the original UTF-8
(composed into a single quasi-character) and they will be written back
correctly as UTF-8, assuming you don't break the composed sequences.
If you read such characters from UTF-16 or UTF-7 data, they are
substituted with the Unicode `replacement character', and you lose
information.

** Accented ISO-8859-1 characters are displayed as | or _.

Try other font set sizes (S-mouse-1).  If the problem persists with
other sizes as well, your text is corrupted, probably through software
that is not 8-bit clean.  If the problem goes away with another font
size, it's probably because some fonts pretend to be ISO-8859-1 fonts
when they are really ASCII fonts. In particular the schumacher-clean
fonts have this bug in some versions of X.

To see what glyphs are included in a font, use `xfd', like this:

  xfd -fn -schumacher-clean-medium-r-normal--12-120-75-75-c-60-iso8859-1

If this shows only ASCII glyphs, the font is indeed the source of the problem.

The solution is to remove the corresponding lines from the appropriate
`fonts.alias' file, then run `mkfontdir' in that directory, and then run
`xset fp rehash'.

** The `oc-unicode' package doesn't work with Emacs 21.

This package tries to define more private charsets than there are free
slots now.  The current built-in Unicode support is actually more
flexible.  (Use option `utf-translate-cjk-mode' if you need CJK
support.)  Files encoded as emacs-mule using oc-unicode aren't
generally read correctly by Emacs 21.

** After a while, Emacs slips into unibyte mode.

The VM mail package, which is not part of Emacs, sometimes does
  (standard-display-european t)
That should be changed to
  (standard-display-european 1 t)

* X runtime problems

** X keyboard problems

*** You "lose characters" after typing Compose Character key.

This is because the Compose Character key is defined as the keysym
Multi_key, and Emacs (seeing that) does the proper X11
character-composition processing.  If you don't want your Compose key
to do that, you can redefine it with xmodmap.

For example, here's one way to turn it into a Meta key:

    xmodmap -e "keysym Multi_key = Meta_L"

If all users at your site of a particular keyboard prefer Meta to
Compose, you can make the remapping happen automatically by adding the
xmodmap command to the xdm setup script for that display.

*** Using X Windows, control-shift-leftbutton makes Emacs hang.

Use the shell command `xset bc' to make the old X Menu package work.

*** C-SPC fails to work on Fedora GNU/Linux (or with fcitx input method).

Fedora Core 4 steals the C-SPC key by default for the `iiimx' program
which is the input method for some languages.  It blocks Emacs users
from using the C-SPC key for `set-mark-command'.

One solutions is to remove the `<Ctrl>space' from the `Iiimx' file
which can be found in the `/usr/lib/X11/app-defaults' directory.
However, that requires root access.

Another is to specify `Emacs*useXIM: false' in your X resources.

Another is to build Emacs with the `--without-xim' configure option.

The same problem happens on any other system if you are using fcitx
(Chinese input method) which by default use C-SPC for toggling.  If
you want to use fcitx with Emacs, you have two choices.  Toggle fcitx
by another key (e.g. C-\) by modifying ~/.fcitx/config, or be
accustomed to use C-@ for `set-mark-command'.

*** M-SPC seems to be ignored as input.

See if your X server is set up to use this as a command
for character composition.

*** The S-C-t key combination doesn't get passed to Emacs on X.

This happens because some X configurations assign the Ctrl-Shift-t
combination the same meaning as the Multi_key.  The offending
definition is in the file `...lib/X11/locale/iso8859-1/Compose'; there
might be other similar combinations which are grabbed by X for similar
purposes.

We think that this can be countermanded with the `xmodmap' utility, if
you want to be able to bind one of these key sequences within Emacs.

*** Under X, C-v and/or other keys don't work.

These may have been intercepted by your window manager.  In
particular, AfterStep 1.6 is reported to steal C-v in its default
configuration.  Various Meta keys are also likely to be taken by the
configuration of the `feel'.  See the WM's documentation for how to
change this.

*** Clicking C-mouse-2 in the scroll bar doesn't split the window.

This currently doesn't work with scroll-bar widgets (and we don't know
a good way of implementing it with widgets).  If Emacs is configured
--without-toolkit-scroll-bars, C-mouse-2 on the scroll bar does work.

*** Inability to send an Alt-modified key, when Emacs is communicating
directly with an X server.

If you have tried to bind an Alt-modified key as a command, and it
does not work to type the command, the first thing you should check is
whether the key is getting through to Emacs.  To do this, type C-h c
followed by the Alt-modified key.  C-h c should say what kind of event
it read.  If it says it read an Alt-modified key, then make sure you
have made the key binding correctly.

If C-h c reports an event that doesn't have the Alt modifier, it may
be because your X server has no key for the Alt modifier.  The X
server that comes from MIT does not set up the Alt modifier by default.

If your keyboard has keys named Alt, you can enable them as follows:

    xmodmap -e 'add mod2 = Alt_L'
    xmodmap -e 'add mod2 = Alt_R'

If the keyboard has just one key named Alt, then only one of those
commands is needed.  The modifier `mod2' is a reasonable choice if you
are using an unmodified MIT version of X.  Otherwise, choose any
modifier bit not otherwise used.

If your keyboard does not have keys named Alt, you can use some other
keys.  Use the keysym command in xmodmap to turn a function key (or
some other 'spare' key) into Alt_L or into Alt_R, and then use the
commands show above to make them modifier keys.

Note that if you have Alt keys but no Meta keys, Emacs translates Alt
into Meta.  This is because of the great importance of Meta in Emacs.

** Window-manager and toolkit-related problems

*** Metacity: Resizing Emacs or ALT-Tab causes X to be unresponsive.

This happens sometimes when using Metacity.  Resizing Emacs or ALT-Tab:bing
makes the system unresponsive to the mouse or the keyboard.  Killing Emacs
or shifting out from X11 and back again usually cures it (i.e. Ctrl-Alt-F1
and then Alt-F7).  A bug for it is here:
https://bugs.launchpad.net/ubuntu/+source/metacity/+bug/231034.
Note that a permanent fix seems to be to disable "assistive technologies".

*** Gnome: Emacs receives input directly from the keyboard, bypassing XIM.

This seems to happen when gnome-settings-daemon version 2.12 or later
is running.  If gnome-settings-daemon is not running, Emacs receives
input through XIM without any problem.  Furthermore, this seems only
to happen in *.UTF-8 locales; zh_CN.GB2312 and zh_CN.GBK locales, for
example, work fine.  A bug report has been filed in the Gnome
bugzilla: http://bugzilla.gnome.org/show_bug.cgi?id=357032

*** Gnome: Emacs' xterm-mouse-mode doesn't work on the Gnome terminal.

A symptom of this bug is that double-clicks insert a control sequence
into the buffer.  The reason this happens is an apparent
incompatibility of the Gnome terminal with Xterm, which also affects
other programs using the Xterm mouse interface.  A problem report has
been filed.

*** KDE: When running on KDE, colors or fonts are not as specified for Emacs,
or messed up.

For example, you could see background you set for Emacs only in the
empty portions of the Emacs display, while characters have some other
background.

This happens because KDE's defaults apply its color and font
definitions even to applications that weren't compiled for KDE.  The
solution is to uncheck the "Apply fonts and colors to non-KDE apps"
option in Preferences->Look&Feel->Style (KDE 2).  In KDE 3, this option
is in the "Colors" section, rather than "Style".

Alternatively, if you do want the KDE defaults to apply to other
applications, but not to Emacs, you could modify the file `Emacs.ad'
(should be in the `/usr/share/apps/kdisplay/app-defaults/' directory)
so that it doesn't set the default background and foreground only for
Emacs.  For example, make sure the following resources are either not
present or commented out:

   Emacs.default.attributeForeground
   Emacs.default.attributeBackground
   Emacs*Foreground
   Emacs*Background

It is also reported that a bug in the gtk-engines-qt engine can cause this if
Emacs is compiled with Gtk+.
The bug is fixed in version 0.7 or newer of gtk-engines-qt.

*** KDE: Emacs hangs on KDE when a large portion of text is killed.

This is caused by a bug in the KDE applet `klipper' which periodically
requests the X clipboard contents from applications.  Early versions
of klipper don't implement the ICCCM protocol for large selections,
which leads to Emacs being flooded with selection requests.  After a
while, Emacs may print a message:

  Timed out waiting for property-notify event

A workaround is to not use `klipper'.  An upgrade to the `klipper' that
comes with KDE 3.3 or later also solves the problem.

*** CDE: Frames may cover dialogs they created when using CDE.

This can happen if you have "Allow Primary Windows On Top" enabled which
seems to be the default in the Common Desktop Environment.
To change, go in to "Desktop Controls" -> "Window Style Manager"
and uncheck "Allow Primary Windows On Top".

*** Xaw3d : When using Xaw3d scroll bars without arrows, the very first mouse
click in a scroll bar might be ignored by the scroll bar widget.  This
is probably a bug in Xaw3d; when Xaw3d is compiled with arrows, the
problem disappears.

*** Xaw: There are known binary incompatibilities between Xaw, Xaw3d, neXtaw,
XawM and the few other derivatives of Xaw.  So when you compile with
one of these, it may not work to dynamically link with another one.
For example, strange problems, such as Emacs exiting when you type
"C-x 1", were reported when Emacs compiled with Xaw3d and libXaw was
used with neXtaw at run time.

The solution is to rebuild Emacs with the toolkit version you actually
want to use, or set LD_PRELOAD to preload the same toolkit version you
built Emacs with.

*** Open Motif: Problems with file dialogs in Emacs built with Open Motif.

When Emacs 21 is built with Open Motif 2.1, it can happen that the
graphical file dialog boxes do not work properly.  The "OK", "Filter"
and "Cancel" buttons do not respond to mouse clicks.  Dragging the
file dialog window usually causes the buttons to work again.

The solution is to use LessTif instead.  LessTif is a free replacement
for Motif.  See the file INSTALL for information on how to do this.

Another workaround is not to use the mouse to trigger file prompts,
but to use the keyboard.  This way, you will be prompted for a file in
the minibuffer instead of a graphical file dialog.

*** LessTif: Problems in Emacs built with LessTif.

The problems seem to depend on the version of LessTif and the Motif
emulation for which it is set up.

Only the Motif 1.2 emulation seems to be stable enough in LessTif.
LessTif 0.92-17's Motif 1.2 emulation seems to work okay on FreeBSD.
On GNU/Linux systems, lesstif-0.92.6 configured with "./configure
--enable-build-12 --enable-default-12" is reported to be the most
successful.  The binary GNU/Linux package
lesstif-devel-0.92.0-1.i386.rpm was reported to have problems with
menu placement.

On some systems, even with Motif 1.2 emulation, Emacs occasionally
locks up, grabbing all mouse and keyboard events.  We still don't know
what causes these problems; they are not reproducible by Emacs developers.

*** Motif: The Motif version of Emacs paints the screen a solid color.

This has been observed to result from the following X resource:

   Emacs*default.attributeFont: -*-courier-medium-r-*-*-*-140-*-*-*-*-iso8859-*

That the resource has this effect indicates a bug in something, but we
do not yet know what.  If it is an Emacs bug, we hope someone can
explain what the bug is so we can fix it.  In the mean time, removing
the resource prevents the problem.

** General X problems

*** Redisplay using X11 is much slower than previous Emacs versions.

We've noticed that certain X servers draw the text much slower when
scroll bars are on the left.  We don't know why this happens.  If this
happens to you, you can work around it by putting the scroll bars
on the right (as they were in Emacs 19).

Here's how to do this:

  (set-scroll-bar-mode 'right)

If you're not sure whether (or how much) this problem affects you,
try that and see how much difference it makes.  To set things back
to normal, do

  (set-scroll-bar-mode 'left)

*** Error messages about undefined colors on X.

The messages might say something like this:

   Unable to load color "grey95"

(typically, in the `*Messages*' buffer), or something like this:

  Error while displaying tooltip: (error Undefined color lightyellow)

These problems could happen if some other X program has used up too
many colors of the X palette, leaving Emacs with insufficient system
resources to load all the colors it needs.

A solution is to exit the offending X programs before starting Emacs.

"undefined color" messages can also occur if the RgbPath entry in the
X configuration file is incorrect, or the rgb.txt file is not where
X expects to find it.

*** Improving performance with slow X connections.

There are several ways to improve this performance, any subset of which can
be carried out at the same time:

1) If you don't need X Input Methods (XIM) for entering text in some
   language you use, you can improve performance on WAN links by using
   the X resource useXIM to turn off use of XIM.  This does not affect
   the use of Emacs' own input methods, which are part of the Leim
   package.

2) If the connection is very slow, you might also want to consider
   switching off scroll bars, menu bar, and tool bar.  Adding the
   following forms to your .emacs file will accomplish that, but only
   after the the initial frame is displayed:

    (scroll-bar-mode -1)
    (menu-bar-mode -1)
    (tool-bar-mode -1)

   For still quicker startup, put these X resources in your .Xdefaults
   file:

    Emacs.verticalScrollBars: off
    Emacs.menuBar: off
    Emacs.toolBar: off

3) Use ssh to forward the X connection, and enable compression on this
   forwarded X connection (ssh -XC remotehostname emacs ...).

4) Use lbxproxy on the remote end of the connection.  This is an interface
   to the low bandwidth X extension in most modern X servers, which
   improves performance dramatically, at the slight expense of correctness
   of the X protocol.  lbxproxy acheives the performance gain by grouping
   several X requests in one TCP packet and sending them off together,
   instead of requiring a round-trip for each X request in a separate
   packet.  The switches that seem to work best for emacs are:
    -noatomsfile  -nowinattr  -cheaterrors -cheatevents
   Note that the -nograbcmap option is known to cause problems.
   For more about lbxproxy, see:
   http://www.xfree86.org/4.3.0/lbxproxy.1.html

5) If copying and killing is slow, try to disable the interaction with the
   native system's clipboard by adding these lines to your .emacs file:
     (setq interprogram-cut-function nil)
     (setq interprogram-paste-function nil)

*** Emacs gives the error, Couldn't find per display information.

This can result if the X server runs out of memory because Emacs uses
a large number of fonts.  On systems where this happens, C-h h is
likely to cause it.

We do not know of a way to prevent the problem.

*** Emacs does not notice when you release the mouse.

There are reports that this happened with (some) Microsoft mice and
that replacing the mouse made it stop.

*** You can't select from submenus (in the X toolkit version).

On certain systems, mouse-tracking and selection in top-level menus
works properly with the X toolkit, but neither of them works when you
bring up a submenu (such as Bookmarks or Compare or Apply Patch, in
the Files menu).

This works on most systems.  There is speculation that the failure is
due to bugs in old versions of X toolkit libraries, but no one really
knows.  If someone debugs this and finds the precise cause, perhaps a
workaround can be found.

*** An error message such as `X protocol error: BadMatch (invalid
parameter attributes) on protocol request 93'.

This comes from having an invalid X resource, such as
   emacs*Cursor:   black
(which is invalid because it specifies a color name for something
that isn't a color.)

The fix is to correct your X resources.

*** Slow startup on X11R6 with X windows.

If Emacs takes two minutes to start up on X11R6, see if your X
resources specify any Adobe fonts.  That causes the type-1 font
renderer to start up, even if the font you asked for is not a type-1
font.

One way to avoid this problem is to eliminate the type-1 fonts from
your font path, like this:

        xset -fp /usr/X11R6/lib/X11/fonts/Type1/

*** Pull-down menus appear in the wrong place, in the toolkit version of Emacs.

An X resource of this form can cause the problem:

   Emacs*geometry:      80x55+0+0

This resource is supposed to apply, and does apply, to the menus
individually as well as to Emacs frames.  If that is not what you
want, rewrite the resource.

To check thoroughly for such resource specifications, use `xrdb
-query' to see what resources the X server records, and also look at
the user's ~/.Xdefaults and ~/.Xdefaults-* files.

*** Emacs running under X Windows does not handle mouse clicks.
*** `emacs -geometry 80x20' finds a file named `80x20'.

One cause of such problems is having (setq term-file-prefix nil) in
your .emacs file.  Another cause is a bad value of EMACSLOADPATH in
the environment.

*** X Windows doesn't work if DISPLAY uses a hostname.

People have reported kernel bugs in certain systems that cause Emacs
not to work with X Windows if DISPLAY is set using a host name.  But
the problem does not occur if DISPLAY is set to `unix:0.0'.  I think
the bug has to do with SIGIO or FIONREAD.

You may be able to compensate for the bug by doing (set-input-mode nil nil).
However, that has the disadvantage of turning off interrupts, so that
you are unable to quit out of a Lisp program by typing C-g.

*** Prevent double pastes in X

The problem:  a region, such as a command, is pasted twice when you copy
it with your mouse from GNU Emacs to an xterm or an RXVT shell in X.
The solution:  try the following in your X configuration file,
/etc/X11/xorg.conf  This should enable both PS/2 and USB mice for
single copies.  You do not need any other drivers or options.

    Section "InputDevice"
            Identifier  "Generic Mouse"
            Driver      "mousedev"
            Option      "Device"           "/dev/input/mice"
    EndSection

* Runtime problems on character terminals

** The meta key does not work on xterm.
Typing M-x rings the terminal bell, and inserts a string like ";120~".
For recent xterm versions (>= 216), Emacs uses xterm's modifyOtherKeys
feature to generate strings for key combinations that are not
otherwise usable.  One circumstance in which this can cause problems
is if you have specified the X resource

  xterm*VT100.Translations

to contain translations that use the meta key.  Then xterm will not
use meta in modified function-keys, which confuses Emacs.  To fix
this, you can remove the X resource or put this in your init file:

  (xterm-remove-modify-other-keys)

** Emacs spontaneously displays "I-search: " at the bottom of the screen.

This means that Control-S/Control-Q (XON/XOFF) "flow control" is being
used.  C-s/C-q flow control is bad for Emacs editors because it takes
away C-s and C-q as user commands.  Since editors do not output long
streams of text without user commands, there is no need for a
user-issuable "stop output" command in an editor; therefore, a
properly designed flow control mechanism would transmit all possible
input characters without interference.  Designing such a mechanism is
easy, for a person with at least half a brain.

There are three possible reasons why flow control could be taking place:

  1) Terminal has not been told to disable flow control
  2) Insufficient padding for the terminal in use
  3) Some sort of terminal concentrator or line switch is responsible

First of all, many terminals have a set-up mode which controls whether
they generate XON/XOFF flow control characters.  This must be set to
"no XON/XOFF" in order for Emacs to work.  (For example, on a VT220
you may select "No XOFF" in the setup menu.)  Sometimes there is an
escape sequence that the computer can send to turn flow control off
and on.  If so, perhaps the termcap `ti' string should turn flow
control off, and the `te' string should turn it on.

Once the terminal has been told "no flow control", you may find it
needs more padding.  The amount of padding Emacs sends is controlled
by the termcap entry for the terminal in use, and by the output baud
rate as known by the kernel.  The shell command `stty' will print
your output baud rate; `stty' with suitable arguments will set it if
it is wrong.  Setting to a higher speed causes increased padding.  If
the results are wrong for the correct speed, there is probably a
problem in the termcap entry.  You must speak to a local Unix wizard
to fix this.  Perhaps you are just using the wrong terminal type.

For terminals that lack a "no flow control" mode, sometimes just
giving lots of padding will prevent actual generation of flow control
codes.  You might as well try it.

If you are really unlucky, your terminal is connected to the computer
through a concentrator which sends XON/XOFF flow control to the
computer, or it insists on sending flow control itself no matter how
much padding you give it.  Unless you can figure out how to turn flow
control off on this concentrator (again, refer to your local wizard),
you are screwed!  You should have the terminal or concentrator
replaced with a properly designed one.  In the mean time, some drastic
measures can make Emacs semi-work.

You can make Emacs ignore C-s and C-q and let the operating system
handle them.  To do this on a per-session basis, just type M-x
enable-flow-control RET.  You will see a message that C-\ and C-^ are
now translated to C-s and C-q.  (Use the same command M-x
enable-flow-control to turn *off* this special mode.  It toggles flow
control handling.)

If C-\ and C-^ are inconvenient for you (for example, if one of them
is the escape character of your terminal concentrator), you can choose
other characters by setting the variables flow-control-c-s-replacement
and flow-control-c-q-replacement.  But choose carefully, since all
other control characters are already used by emacs.

IMPORTANT: if you type C-s by accident while flow control is enabled,
Emacs output will freeze, and you will have to remember to type C-q in
order to continue.

If you work in an environment where a majority of terminals of a
certain type are flow control hobbled, you can use the function
`enable-flow-control-on' to turn on this flow control avoidance scheme
automatically.  Here is an example:

(enable-flow-control-on "vt200" "vt300" "vt101" "vt131")

If this isn't quite correct (e.g. you have a mixture of flow-control hobbled
and good vt200 terminals), you can still run enable-flow-control
manually.

I have no intention of ever redesigning the Emacs command set for the
assumption that terminals use C-s/C-q flow control.  XON/XOFF flow
control technique is a bad design, and terminals that need it are bad
merchandise and should not be purchased.  Now that X is becoming
widespread, XON/XOFF seems to be on the way out.  If you can get some
use out of GNU Emacs on inferior terminals, more power to you, but I
will not make Emacs worse for properly designed systems for the sake
of inferior systems.

** Control-S and Control-Q commands are ignored completely.

For some reason, your system is using brain-damaged C-s/C-q flow
control despite Emacs's attempts to turn it off.  Perhaps your
terminal is connected to the computer through a concentrator
that wants to use flow control.

You should first try to tell the concentrator not to use flow control.
If you succeed in this, try making the terminal work without
flow control, as described in the preceding section.

If that line of approach is not successful, map some other characters
into C-s and C-q using keyboard-translate-table.  The example above
shows how to do this with C-^ and C-\.

** Screen is updated wrong, but only on one kind of terminal.

This could mean that the termcap entry you are using for that
terminal is wrong, or it could mean that Emacs has a bug handing
the combination of features specified for that terminal.

The first step in tracking this down is to record what characters
Emacs is sending to the terminal.  Execute the Lisp expression
(open-termscript "./emacs-script") to make Emacs write all
terminal output into the file ~/emacs-script as well; then do
what makes the screen update wrong, and look at the file
and decode the characters using the manual for the terminal.
There are several possibilities:

1) The characters sent are correct, according to the terminal manual.

In this case, there is no obvious bug in Emacs, and most likely you
need more padding, or possibly the terminal manual is wrong.

2) The characters sent are incorrect, due to an obscure aspect
 of the terminal behavior not described in an obvious way by termcap.

This case is hard.  It will be necessary to think of a way for
Emacs to distinguish between terminals with this kind of behavior
and other terminals that behave subtly differently but are
classified the same by termcap; or else find an algorithm for
Emacs to use that avoids the difference.  Such changes must be
tested on many kinds of terminals.

3) The termcap entry is wrong.

See the file etc/TERMS for information on changes
that are known to be needed in commonly used termcap entries
for certain terminals.

4) The characters sent are incorrect, and clearly cannot be
 right for any terminal with the termcap entry you were using.

This is unambiguously an Emacs bug, and can probably be fixed
in termcap.c, tparam.c, term.c, scroll.c, cm.c or dispnew.c.

** Control-S and Control-Q commands are ignored completely on a net connection.

Some versions of rlogin (and possibly telnet) do not pass flow
control characters to the remote system to which they connect.
On such systems, emacs on the remote system cannot disable flow
control on the local system.  Sometimes `rlogin -8' will avoid this problem.

One way to cure this is to disable flow control on the local host
(the one running rlogin, not the one running rlogind) using the
stty command, before starting the rlogin process.  On many systems,
"stty start u stop u" will do this.  On some systems, use
"stty -ixon" instead.

Some versions of tcsh will prevent even this from working.  One way
around this is to start another shell before starting rlogin, and
issue the stty command to disable flow control from that shell.

If none of these methods work, the best solution is to type
M-x enable-flow-control at the beginning of your emacs session, or
if you expect the problem to continue, add a line such as the
following to your .emacs (on the host running rlogind):

(enable-flow-control-on "vt200" "vt300" "vt101" "vt131")

See the entry about spontaneous display of I-search (above) for more info.

** Output from Control-V is slow.

On many bit-map terminals, scrolling operations are fairly slow.
Often the termcap entry for the type of terminal in use fails
to inform Emacs of this.  The two lines at the bottom of the screen
before a Control-V command are supposed to appear at the top after
the Control-V command.  If Emacs thinks scrolling the lines is fast,
it will scroll them to the top of the screen.

If scrolling is slow but Emacs thinks it is fast, the usual reason is
that the termcap entry for the terminal you are using does not
specify any padding time for the `al' and `dl' strings.  Emacs
concludes that these operations take only as much time as it takes to
send the commands at whatever line speed you are using.  You must
fix the termcap entry to specify, for the `al' and `dl', as much
time as the operations really take.

Currently Emacs thinks in terms of serial lines which send characters
at a fixed rate, so that any operation which takes time for the
terminal to execute must also be padded.  With bit-map terminals
operated across networks, often the network provides some sort of
flow control so that padding is never needed no matter how slow
an operation is.  You must still specify a padding time if you want
Emacs to realize that the operation takes a long time.  This will
cause padding characters to be sent unnecessarily, but they do
not really cost much.  They will be transmitted while the scrolling
is happening and then discarded quickly by the terminal.

Most bit-map terminals provide commands for inserting or deleting
multiple lines at once.  Define the `AL' and `DL' strings in the
termcap entry to say how to do these things, and you will have
fast output without wasted padding characters.  These strings should
each contain a single %-spec saying how to send the number of lines
to be scrolled.  These %-specs are like those in the termcap
`cm' string.

You should also define the `IC' and `DC' strings if your terminal
has a command to insert or delete multiple characters.  These
take the number of positions to insert or delete as an argument.

A `cs' string to set the scrolling region will reduce the amount
of motion you see on the screen when part of the screen is scrolled.

** You type Control-H (Backspace) expecting to delete characters.

Put `stty dec' in your .login file and your problems will disappear
after a day or two.

The choice of Backspace for erasure was based on confusion, caused by
the fact that backspacing causes erasure (later, when you type another
character) on most display terminals.  But it is a mistake.  Deletion
of text is not the same thing as backspacing followed by failure to
overprint.  I do not wish to propagate this confusion by conforming
to it.

For this reason, I believe `stty dec' is the right mode to use,
and I have designed Emacs to go with that.  If there were a thousand
other control characters, I would define Control-h to delete as well;
but there are not very many other control characters, and I think
that providing the most mnemonic possible Help character is more
important than adapting to people who don't use `stty dec'.

If you are obstinate about confusing buggy overprinting with deletion,
you can redefine Backspace in your .emacs file:
  (global-set-key "\b" 'delete-backward-char)
You can probably access  help-command  via f1.

** Colors are not available on a tty or in xterm.

Emacs 21 supports colors on character terminals and terminal
emulators, but this support relies on the terminfo or termcap database
entry to specify that the display supports color.  Emacs looks at the
"Co" capability for the terminal to find out how many colors are
supported; it should be non-zero to activate the color support within
Emacs.  (Most color terminals support 8 or 16 colors.)  If your system
uses terminfo, the name of the capability equivalent to "Co" is
"colors".

In addition to the "Co" capability, Emacs needs the "op" (for
``original pair'') capability, which tells how to switch the terminal
back to the default foreground and background colors.  Emacs will not
use colors if this capability is not defined.  If your terminal entry
doesn't provide such a capability, try using the ANSI standard escape
sequence \E[00m (that is, define a new termcap/terminfo entry and make
it use your current terminal's entry plus \E[00m for the "op"
capability).

Finally, the "NC" capability (terminfo name: "ncv") tells Emacs which
attributes cannot be used with colors.  Setting this capability
incorrectly might have the effect of disabling colors; try setting
this capability to `0' (zero) and see if that helps.

Emacs uses the database entry for the terminal whose name is the value
of the environment variable TERM.  With `xterm', a common terminal
entry that supports color is `xterm-color', so setting TERM's value to
`xterm-color' might activate the color support on an xterm-compatible
emulator.

Beginning with version 22.1, Emacs supports the --color command-line
option which may be used to force Emacs to use one of a few popular
modes for getting colors on a tty.  For example, --color=ansi8 sets up
for using the ANSI-standard escape sequences that support 8 colors.

Some modes do not use colors unless you turn on the Font-lock mode.
Some people have long ago set their `~/.emacs' files to turn on
Font-lock on X only, so they won't see colors on a tty.  The
recommended way of turning on Font-lock is by typing "M-x
global-font-lock-mode RET" or by customizing the variable
`global-font-lock-mode'.

* Runtime problems specific to individual Unix variants

** GNU/Linux

*** GNU/Linux: Process output is corrupted.

There is a bug in Linux kernel 2.6.10 PTYs that can cause emacs to
read corrupted process output.

*** GNU/Linux: Remote access to CVS with SSH causes file corruption.

If you access a remote CVS repository via SSH, files may be corrupted
due to bad interaction between CVS, SSH, and libc.

To fix the problem, save the following script into a file, make it
executable, and set CVS_RSH environment variable to the file name of
the script:

#!/bin/bash
exec 2> >(exec cat >&2 2>/dev/null)
exec ssh "$@"

*** GNU/Linux: Truncated svn annotate output with SSH.
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=7791

The symptoms are: you are accessing a svn repository over SSH.
You use vc-annotate on a large (several thousand line) file, and the
result is truncated around the 1000 line mark.  It works fine with
other access methods (eg http), or from outside Emacs.

This may be a similar libc/SSH issue to the one mentioned above for CVS.
A similar workaround seems to be effective: create a script with the
same contents as the one used above for CVS_RSH, and set the SVN_SSH
environment variable to point to it.

*** GNU/Linux: On Linux-based GNU systems using libc versions 5.4.19 through
5.4.22, Emacs crashes at startup with a segmentation fault.

This problem happens if libc defines the symbol __malloc_initialized.
One known solution is to upgrade to a newer libc version.  5.4.33 is
known to work.

*** GNU/Linux: After upgrading to a newer version of Emacs,
the Meta key stops working.

This was reported to happen on a GNU/Linux system distributed by
Mandrake.  The reason is that the previous version of Emacs was
modified by Mandrake to make the Alt key act as the Meta key, on a
keyboard where the Windows key is the one which produces the Meta
modifier.  A user who started using a newer version of Emacs, which
was not hacked by Mandrake, expected the Alt key to continue to act as
Meta, and was astonished when that didn't happen.

The solution is to find out what key on your keyboard produces the Meta
modifier, and use that key instead.  Try all of the keys to the left
and to the right of the space bar, together with the `x' key, and see
which combination produces "M-x" in the echo area.  You can also use
the `xmodmap' utility to show all the keys which produce a Meta
modifier:

         xmodmap -pk | egrep -i "meta|alt"

A more convenient way of finding out which keys produce a Meta modifier
is to use the `xkbprint' utility, if it's available on your system:

         xkbprint 0:0 /tmp/k.ps

This produces a PostScript file `/tmp/k.ps' with a picture of your
keyboard; printing that file on a PostScript printer will show what
keys can serve as Meta.

The `xkeycaps' also shows a visual representation of the current
keyboard settings.  It also allows to modify them.

*** GNU/Linux: slow startup on Linux-based GNU systems.

People using systems based on the Linux kernel sometimes report that
startup takes 10 to 15 seconds longer than `usual'.

This is because Emacs looks up the host name when it starts.
Normally, this takes negligible time; the extra delay is due to
improper system configuration.  This problem can occur for both
networked and non-networked machines.

Here is how to fix the configuration.  It requires being root.

**** Networked Case.

First, make sure the files `/etc/hosts' and `/etc/host.conf' both
exist.  The first line in the `/etc/hosts' file should look like this
(replace HOSTNAME with your host name):

    127.0.0.1      HOSTNAME

Also make sure that the `/etc/host.conf' files contains the following
lines:

    order hosts, bind
    multi on

Any changes, permanent and temporary, to the host name should be
indicated in the `/etc/hosts' file, since it acts a limited local
database of addresses and names (e.g., some SLIP connections
dynamically allocate ip addresses).

**** Non-Networked Case.

The solution described in the networked case applies here as well.
However, if you never intend to network your machine, you can use a
simpler solution: create an empty `/etc/host.conf' file.  The command
`touch /etc/host.conf' suffices to create the file.  The `/etc/hosts'
file is not necessary with this approach.

*** GNU/Linux: Emacs on a tty switches the cursor to large blinking block.

This was reported to happen on some GNU/Linux systems which use
ncurses version 5.0, but could be relevant for other versions as well.
These versions of ncurses come with a `linux' terminfo entry, where
the "cvvis" capability (termcap "vs") is defined as "\E[?25h\E[?8c"
(show cursor, change size).  This escape sequence switches on a
blinking hardware text-mode cursor whose size is a full character
cell.  This blinking cannot be stopped, since a hardware cursor
always blinks.

A work-around is to redefine the "cvvis" capability so that it
enables a *software* cursor.  The software cursor works by inverting
the colors of the character at point, so what you see is a block
cursor that doesn't blink.  For this to work, you need to redefine
the "cnorm" capability as well, so that it operates on the software
cursor instead of the hardware cursor.

To this end, run "infocmp linux > linux-term", edit the file
`linux-term' to make both the "cnorm" and "cvvis" capabilities send
the sequence "\E[?25h\E[?17;0;64c", and then run "tic linux-term" to
produce a modified terminfo entry.

Alternatively, if you want a blinking underscore as your Emacs cursor,
change the "cvvis" capability to send the "\E[?25h\E[?0c" command.

*** GNU/Linux: Error messages `internal facep []' happen on GNU/Linux systems.

There is a report that replacing libc.so.5.0.9 with libc.so.5.2.16
caused this to start happening.  People are not sure why, but the
problem seems unlikely to be in Emacs itself.  Some suspect that it
is actually Xlib which won't work with libc.so.5.2.16.

Using the old library version is a workaround.

** FreeBSD

*** FreeBSD 2.1.5: useless symbolic links remain in /tmp or other
directories that have the +t bit.

This is because of a kernel bug in FreeBSD 2.1.5 (fixed in 2.2).
Emacs uses symbolic links to implement file locks.  In a directory
with +t bit, the directory owner becomes the owner of the symbolic
link, so that it cannot be removed by anyone else.

If you don't like those useless links, you can let Emacs not to using
file lock by adding #undef CLASH_DETECTION to config.h.

*** FreeBSD: Getting a Meta key on the console.

By default, neither Alt nor any other key acts as a Meta key on
FreeBSD, but this can be changed using kbdcontrol(1).  Dump the
current keymap to a file with the command

  $ kbdcontrol -d >emacs.kbd

Edit emacs.kbd, and give the key you want to be the Meta key the
definition `meta'.  For instance, if your keyboard has a ``Windows''
key with scan code 105, change the line for scan code 105 in emacs.kbd
to look like this

  105   meta   meta   meta   meta   meta   meta   meta   meta    O

to make the Windows key the Meta key.  Load the new keymap with

  $ kbdcontrol -l emacs.kbd

** HP-UX

*** HP/UX : Shell mode gives the message, "`tty`: Ambiguous".

christos@theory.tn.cornell.edu says:

The problem is that in your .cshrc you have something that tries to
execute `tty`. If you are not running the shell on a real tty then
tty will print "not a tty". Csh expects one word in some places,
but tty is giving it back 3.

The solution is to add a pair of quotes around `tty` to make it a single
word:

if (`tty` == "/dev/console")

should be changed to:

if ("`tty`" == "/dev/console")

Even better, move things that set up terminal sections out of .cshrc
and into .login.

*** HP/UX: `Pid xxx killed due to text modification or page I/O error'.

On HP/UX, you can get that error when the Emacs executable is on an NFS
file system.  HP/UX responds this way if it tries to swap in a page and
does not get a response from the server within a timeout whose default
value is just ten seconds.

If this happens to you, extend the timeout period.

*** HP/UX: The right Alt key works wrong on German HP keyboards (and perhaps
other non-English HP keyboards too).

This is because HP-UX defines the modifiers wrong in X.  Here is a
shell script to fix the problem; be sure that it is run after VUE
configures the X server.

    xmodmap 2> /dev/null - << EOF
    keysym Alt_L = Meta_L
    keysym Alt_R = Meta_R
    EOF

    xmodmap - << EOF
    clear mod1
    keysym Mode_switch = NoSymbol
    add mod1 = Meta_L
    keysym Meta_R = Mode_switch
    add mod2 = Mode_switch
    EOF

*** HP/UX: "Cannot find callback list" messages from dialog boxes in
Emacs built with Motif.

This problem resulted from a bug in GCC 2.4.5.  Newer GCC versions
such as 2.7.0 fix the problem.

*** HP/UX: Emacs does not recognize the AltGr key.

To fix this, set up a file ~/.dt/sessions/sessionetc with executable
rights, containing this text:

--------------------------------
xmodmap 2> /dev/null - << EOF
keysym Alt_L = Meta_L
keysym Alt_R = Meta_R
EOF

xmodmap - << EOF
clear mod1
keysym Mode_switch = NoSymbol
add mod1 = Meta_L
keysym Meta_R = Mode_switch
add mod2 = Mode_switch
EOF
--------------------------------

*** HP/UX 11.0: Emacs makes HP/UX 11.0 crash.

This is a bug in HPUX; HPUX patch PHKL_16260 is said to fix it.

** AIX

*** AIX: Trouble using ptys.

People often install the pty devices on AIX incorrectly.
Use `smit pty' to reinstall them properly.

*** AIXterm: Your Delete key sends a Backspace to the terminal.

The solution is to include in your .Xdefaults the lines:

   *aixterm.Translations: #override <Key>BackSpace: string(0x7f)
   aixterm*ttyModes: erase ^?

This makes your Backspace key send DEL (ASCII 127).

*** AIX: If linking fails because libXbsd isn't found, check if you
are compiling with the system's `cc' and CFLAGS containing `-O5'.  If
so, you have hit a compiler bug.  Please make sure to re-configure
Emacs so that it isn't compiled with `-O5'.

*** AIX 4.3.x or 4.4: Compiling fails.

This could happen if you use /bin/c89 as your compiler, instead of
the default `cc'.  /bin/c89 treats certain warnings, such as benign
redefinitions of macros, as errors, and fails the build.  A solution
is to use the default compiler `cc'.

*** AIX 4: Some programs fail when run in a Shell buffer
with an error message like   No terminfo entry for "unknown".

On AIX, many terminal type definitions are not installed by default.
`unknown' is one of them.  Install the "Special Generic Terminal
Definitions" to make them defined.

** Solaris

We list bugs in current versions here.  See also the section on legacy
systems.

*** On Solaris, C-x doesn't get through to Emacs when you use the console.

This is a Solaris feature (at least on Intel x86 cpus).  Type C-r
C-r C-t, to toggle whether C-x gets through to Emacs.

*** Problem with remote X server on Suns.

On a Sun, running Emacs on one machine with the X server on another
may not work if you have used the unshared system libraries.  This
is because the unshared libraries fail to use YP for host name lookup.
As a result, the host name you specify may not be recognized.

*** Solaris 2.6: Emacs crashes with SIGBUS or SIGSEGV on Solaris after you delete a frame.

We suspect that this is a bug in the X libraries provided by
Sun.  There is a report that one of these patches fixes the bug and
makes the problem stop:

105216-01 105393-01 105518-01 105621-01 105665-01 105615-02 105216-02
105667-01 105401-08 105615-03 105621-02 105686-02 105736-01 105755-03
106033-01 105379-01 105786-01 105181-04 105379-03 105786-04 105845-01
105284-05 105669-02 105837-01 105837-02 105558-01 106125-02 105407-01

Another person using a newer system (kernel patch level Generic_105181-06)
suspects that the bug was fixed by one of these more recent patches:

106040-07  SunOS 5.6: X Input & Output Method patch
106222-01  OpenWindows 3.6: filemgr (ff.core) fixes
105284-12  Motif 1.2.7: sparc Runtime library patch

*** Solaris 7 or 8: Emacs reports a BadAtom error (from X)

This happens when Emacs was built on some other version of Solaris.
Rebuild it on Solaris 8.

*** When using M-x dbx with the SparcWorks debugger, the `up' and `down'
commands do not move the arrow in Emacs.

You can fix this by adding the following line to `~/.dbxinit':

 dbxenv output_short_file_name off

*** On Solaris, CTRL-t is ignored by Emacs when you use
the fr.ISO-8859-15 locale (and maybe other related locales).

You can fix this by editing the file:

        /usr/openwin/lib/locale/iso8859-15/Compose

Near the bottom there is a line that reads:

        Ctrl<t> <quotedbl> <Y>                  : "\276"        threequarters

that should read:

        Ctrl<T> <quotedbl> <Y>                  : "\276"        threequarters

Note the lower case <t>.  Changing this line should make C-t work.

*** On Solaris, Emacs fails to set menu-bar-update-hook on startup, with error
"Error in menu-bar-update-hook: (error Point before start of properties)".
This seems to be a GCC optimization bug that occurs for GCC 4.1.2 (-g
and -g -O2) and GCC 4.2.3 (-g -O and -g -O2).  You can fix this by
compiling with GCC 4.2.3 or CC 5.7, with no optimizations.

** Irix

*** Irix 6.5: Emacs crashes on the SGI R10K, when compiled with GCC.

This seems to be fixed in GCC 2.95.

*** Irix: Trouble using ptys, or running out of ptys.

The program mkpts (which may be in `/usr/adm' or `/usr/sbin') needs to
be set-UID to root, or non-root programs like Emacs will not be able
to allocate ptys reliably.

* Runtime problems specific to MS-Windows

** PATH can contain unexpanded environment variables

Old releases of TCC (version 9) and 4NT (up to version 8) do not correctly
expand App Paths entries of type REG_EXPAND_SZ.  When Emacs is run from TCC
and such an entry exists for emacs.exe, exec-path will contain the
unexpanded entry.  This has been fixed in TCC 10.  For more information,
see bug#2062.

** Setting w32-pass-rwindow-to-system and w32-pass-lwindow-to-system to nil
does not prevent the Start menu from popping up when the left or right
``Windows'' key is pressed.

This was reported to happen when XKeymacs is installed.  At least with
XKeymacs Version 3.47, deactivating XKeymacs when Emacs is active is
not enough to avoid its messing with the keyboard input.  Exiting
XKeymacs completely is reported to solve the problem.

** Windows 95 and networking.

To support server sockets, Emacs 22.1 loads ws2_32.dll.  If this file
is missing, all Emacs networking features are disabled.

Old versions of Windows 95 may not have the required DLL.  To use
Emacs' networking features on Windows 95, you must install the
"Windows Socket 2" update available from MicroSoft's support Web.

** Emacs exits with "X protocol error" when run with an X server for MS-Windows.

A certain X server for Windows had a bug which caused this.
Supposedly the newer 32-bit version of this server doesn't have the
problem.

** Emacs crashes when opening a file with a UNC path and rails-mode is loaded.

Loading rails-mode seems to interfere with UNC path handling. This has been
reported as a bug against both Emacs and rails-mode, so look for an updated
rails-mode that avoids this crash, or avoid using UNC paths if using
rails-mode.

** Known problems with the MS-Windows port of Emacs 22.3

M-x term does not work on MS-Windows.  TTY emulation on Windows is
undocumented, and programs such as stty which are used on posix platforms
to control tty emulation do not exist for native windows terminals.

Using create-fontset-from-ascii-font or the --font startup parameter
with a Chinese, Japanese or Korean font leads to display problems.
Use a Latin-only font as your default font. If you want control over
which font is used to display Chinese, Japanese or Korean character,
use create-fontset-from-fontset-spec to define a fontset.

Frames are not refreshed while the File or Font dialog or a pop-up menu
is displayed. This also means help text for pop-up menus is not
displayed at all.  This is because message handling under Windows is
synchronous, so we cannot handle repaint (or any other) messages while
waiting for a system function to return the result of the dialog or
pop-up menu interaction.

Windows 95 and Windows NT up to version 4.0 do not support help text
for menus.  Help text is only available in later versions of Windows.

When "ClearType" method is selected as the "method to smooth edges of
screen fonts" (in Display Properties, Appearance tab, under
"Effects"), there are various problems related to display of
characters:  Bold fonts can be hard to read, small portions of some
characters could appear chopped, etc.  This happens because, under
ClearType, characters are drawn outside their advertised bounding box.
Emacs 21 disabled the use of ClearType, whereas Emacs 22 allows it and
has some code to enlarge the width of the bounding box.  Apparently,
this display feature needs more changes to get it 100% right.  A
workaround is to disable ClearType.

There are problems with display if mouse-tracking is enabled and the
mouse is moved off a frame, over another frame then back over the first
frame.  A workaround is to click the left mouse button inside the frame
after moving back into it.

Some minor flickering still persists during mouse-tracking, although
not as severely as in 21.1.

An inactive cursor remains in an active window after the Windows
Manager driven switch of the focus, until a key is pressed.

Windows input methods are not recognized by Emacs.  However, some
of these input methods cause the keyboard to send characters encoded
in the appropriate coding system (e.g., ISO 8859-1 for Latin-1
characters, ISO 8859-8 for Hebrew characters, etc.).  To make these
input methods work with Emacs, set the keyboard coding system to the
appropriate value after you activate the Windows input method.  For
example, if you activate the Hebrew input method, type this:

   C-x RET k hebrew-iso-8bit RET

(Emacs ought to recognize the Windows language-change event and set up
the appropriate keyboard encoding automatically, but it doesn't do
that yet.)  In addition, to use these Windows input methods, you
should set your "Language for non-Unicode programs" (on Windows XP,
this is on the Advanced tab of Regional Settings) to the language of
the input method.

To bind keys that produce non-ASCII characters with modifiers, you
must specify raw byte codes. For instance, if you want to bind
META-a-grave to a command, you need to specify this in your `~/.emacs':

  (global-set-key [?\M-\340] ...)

The above example is for the Latin-1 environment where the byte code
of the encoded a-grave is 340 octal.  For other environments, use the
encoding appropriate to that environment.

The %b specifier for format-time-string does not produce abbreviated
month names with consistent widths for some locales on some versions
of Windows. This is caused by a deficiency in the underlying system
library function.

The function set-time-zone-rule gives incorrect results for many
non-US timezones.  This is due to over-simplistic handling of
daylight savings switchovers by the Windows libraries.

Files larger than 4GB cause overflow in the size (represented as a
32-bit integer) reported by `file-attributes'.  This affects Dired as
well, since the Windows port uses a Lisp emulation of `ls' that relies
on `file-attributes'.

Sound playing is not supported with the `:data DATA' key-value pair.
You _must_ use the `:file FILE' method.

** Typing Alt-Shift has strange effects on MS-Windows.

This combination of keys is a command to change keyboard layout.  If
you proceed to type another non-modifier key before you let go of Alt
and Shift, the Alt and Shift act as modifiers in the usual way.  A
more permanent work around is to change it to another key combination,
or disable it in the "Regional and Language Options" applet of the
Control Panel.  (The exact sequence of mouse clicks in the "Regional
and Language Options" applet needed to find the key combination that
changes the keyboard layout depends on your Windows version; for XP,
in the Languages tab, click "Details" and then "Key Settings".)

** Interrupting Cygwin port of Bash from Emacs doesn't work.

Cygwin 1.x builds of the ported Bash cannot be interrupted from the
MS-Windows version of Emacs.  This is due to some change in the Bash
port or in the Cygwin library which apparently make Bash ignore the
keyboard interrupt event sent by Emacs to Bash.  (Older Cygwin ports
of Bash, up to b20.1, did receive SIGINT from Emacs.)

** Accessing remote files with ange-ftp hangs the MS-Windows version of Emacs.

If the FTP client is the Cygwin port of GNU `ftp', this appears to be
due to some bug in the Cygwin DLL or some incompatibility between it
and the implementation of asynchronous subprocesses in the Windows
port of Emacs.  Specifically, some parts of the FTP server responses
are not flushed out, apparently due to buffering issues, which
confuses ange-ftp.

The solution is to downgrade to an older version of the Cygwin DLL
(version 1.3.2 was reported to solve the problem), or use the stock
Windows FTP client, usually found in the `C:\WINDOWS' or 'C:\WINNT'
directory.  To force ange-ftp use the stock Windows client, set the
variable `ange-ftp-ftp-program-name' to the absolute file name of the
client's executable.  For example:

 (setq ange-ftp-ftp-program-name "c:/windows/ftp.exe")

If you want to stick with the Cygwin FTP client, you can work around
this problem by putting this in your `.emacs' file:

 (setq ange-ftp-ftp-program-args '("-i" "-n" "-g" "-v" "--prompt" "")

** lpr commands don't work on MS-Windows with some cheap printers.

This problem may also strike other platforms, but the solution is
likely to be a global one, and not Emacs specific.

Many cheap inkjet, and even some cheap laser printers, do not
print plain text anymore, they will only print through graphical
printer drivers. A workaround on MS-Windows is to use Windows' basic
built in editor to print (this is possibly the only useful purpose it
has):

(setq printer-name "")         ;; notepad takes the default
(setq lpr-command "notepad")   ;; notepad
(setq lpr-switches nil)        ;; not needed
(setq lpr-printer-switch "/P") ;; run notepad as batch printer

** Antivirus software interacts badly with the MS-Windows version of Emacs.

The usual manifestation of these problems is that subprocesses don't
work or even wedge the entire system.  In particular, "M-x shell RET"
was reported to fail to work.  But other commands also sometimes don't
work when an antivirus package is installed.

The solution is to switch the antivirus software to a less aggressive
mode (e.g., disable the ``auto-protect'' feature), or even uninstall
or disable it entirely.

** Pressing the mouse button on MS-Windows does not give a mouse-2 event.

This is usually a problem with the mouse driver. Because most Windows
programs do not do anything useful with the middle mouse button, many
mouse drivers allow you to define the wheel press to do something
different. Some drivers do not even have the option to generate a
middle button press. In such cases, setting the wheel press to
"scroll" sometimes works if you press the button twice. Trying a
generic mouse driver might help.

** Scrolling the mouse wheel on MS-Windows always scrolls the top window.

This is another common problem with mouse drivers. Instead of
generating scroll events, some mouse drivers try to fake scroll bar
movement. But they are not intelligent enough to handle multiple
scroll bars within a frame. Trying a generic mouse driver might help.

** Mail sent through Microsoft Exchange in some encodings appears to be
mangled and is not seen correctly in Rmail or Gnus.  We don't know
exactly what happens, but it isn't an Emacs problem in cases we've
seen.

** On MS-Windows, you cannot use the right-hand ALT key and the left-hand
CTRL key together to type a Control-Meta character.

This is a consequence of a misfeature beyond Emacs's control.

Under Windows, the AltGr key on international keyboards generates key
events with the modifiers Right-Alt and Left-Ctrl.  Since Emacs cannot
distinguish AltGr from an explicit Right-Alt and Left-Ctrl
combination, whenever it sees Right-Alt and Left-Ctrl it assumes that
AltGr has been pressed.  The variable `w32-recognize-altgr' can be set
to nil to tell Emacs that AltGr is really Ctrl and Alt.

** Under some X-servers running on MS-Windows, Emacs' display is incorrect.

The symptoms are that Emacs does not completely erase blank areas of the
screen during scrolling or some other screen operations (e.g., selective
display or when killing a region).  M-x recenter will cause the screen
to be completely redisplayed and the "extra" characters will disappear.

This is known to occur under Exceed 6, and possibly earlier versions
as well; it is reportedly solved in version 6.2.0.16 and later.  The
problem lies in the X-server settings.

There are reports that you can solve the problem with Exceed by
running `Xconfig' from within NT, choosing "X selection", then
un-checking the boxes "auto-copy X selection" and "auto-paste to X
selection".

Of this does not work, please inform bug-gnu-emacs@gnu.org.  Then
please call support for your X-server and see if you can get a fix.
If you do, please send it to bug-gnu-emacs@gnu.org so we can list it here.

* Build-time problems

** Configuration

*** The `configure' script doesn't find the jpeg library.

There are reports that this happens on some systems because the linker
by default only looks for shared libraries, but jpeg distribution by
default only installs a nonshared version of the library, `libjpeg.a'.

If this is the problem, you can configure the jpeg library with the
`--enable-shared' option and then rebuild libjpeg.  This produces a
shared version of libjpeg, which you need to install.  Finally, rerun
the Emacs configure script, which should now find the jpeg library.
Alternatively, modify the generated src/Makefile to link the .a file
explicitly, and edit src/config.h to define HAVE_JPEG.

*** `configure' warns ``accepted by the compiler, rejected by the preprocessor''.

This indicates a mismatch between the C compiler and preprocessor that
configure is using.  For example, on Solaris 10 trying to use
CC=/opt/SUNWspro/bin/cc (the Sun Studio compiler) together with
CPP=/usr/ccs/lib/cpp can result in errors of this form (you may also
see the error ``"/usr/include/sys/isa_defs.h", line 500: undefined control'').

The solution is to tell configure to use the correct C preprocessor
for your C compiler (CPP="/opt/SUNWspro/bin/cc -E" in the above
example).

** Compilation

*** Building Emacs over NFS fails with ``Text file busy''.

This was reported to happen when building Emacs on a GNU/Linux system
(Red Hat Linux 6.2) using a build directory automounted from Solaris
(SunOS 5.6) file server, but it might not be limited to that
configuration alone.  Presumably, the NFS server doesn't commit the
files' data to disk quickly enough, and the Emacs executable file is
left ``busy'' for several seconds after Emacs has finished dumping
itself.  This causes the subsequent commands which invoke the dumped
Emacs executable to fail with the above message.

In some of these cases, a time skew between the NFS server and the
machine where Emacs is built is detected and reported by GNU Make
(it says that some of the files have modification time in the future).
This might be a symptom of NFS-related problems.

If the NFS server runs on Solaris, apply the Solaris patch 105379-05
(Sunos 5.6: /kernel/misc/nfssrv patch).  If that doesn't work, or if
you have a different version of the OS or the NFS server, you can
force the NFS server to use 1KB blocks, which was reported to fix the
problem albeit at a price of slowing down file I/O.  You can force 1KB
blocks by specifying the "-o  rsize=1024,wsize=1024" options to the
`mount' command, or by adding ",rsize=1024,wsize=1024" to the mount
options in the appropriate system configuration file, such as
`/etc/auto.home'.

Alternatively, when Make fails due to this problem, you could wait for
a few seconds and then invoke Make again.  In one particular case,
waiting for 10 or more seconds between the two Make invocations seemed
to work around the problem.

Similar problems can happen if your machine NFS-mounts a directory
onto itself.  Suppose the Emacs sources live in `/usr/local/src' and
you are working on the host called `marvin'.  Then an entry in the
`/etc/fstab' file like the following is asking for trouble:

    marvin:/usr/local/src /usr/local/src ...options.omitted...

The solution is to remove this line from `etc/fstab'.

*** Building a 32-bit executable on a 64-bit GNU/Linux architecture.

First ensure that the necessary 32-bit system libraries and include
files are installed. Then use:

  env CC="gcc -m32" ./configure --build=i386-linux-gnu \
    --x-libraries=/usr/X11R6/lib

(using the location of the 32-bit X libraries on your system).

*** Building Emacs for Cygwin can fail with GCC 3

As of Emacs 22.1, there have been stability problems with Cygwin
builds of Emacs using GCC 3.  Cygwin users are advised to use GCC 4.

*** Building Emacs 23.3 and later will fail under Cygwin 1.5.19

This is a consequence of a change to src/dired.c on 2010-07-27.  The
issue is that Cygwin 1.5.19 did not have d_ino in 'struct dirent'.
See

  http://lists.gnu.org/archive/html/emacs-devel/2010-07/msg01266.html

*** Building the native MS-Windows port fails due to unresolved externals

The linker error messages look like this:

 oo-spd/i386/ctags.o:ctags.c:(.text+0x156e): undefined reference to `_imp__re_set_syntax'
 collect2: ld returned 1 exit status

This happens because GCC finds an incompatible header regex.h
somewhere on the include path, before the version of regex.h supplied
with Emacs.  One such incompatible version of regex.h is part of the
GnuWin32 Regex package.

The solution is to remove the incompatible regex.h from the include
path, when compiling Emacs.  Alternatively, re-run the configure.bat
script with the "-isystem C:/GnuWin32/include" switch (adapt for your
system's place where you keep the GnuWin32 include files) -- this will
cause the compiler to search headers in the directories specified by
the Emacs Makefile _before_ it looks in the GnuWin32 include
directories.

*** Building the native MS-Windows port with Cygwin GCC can fail.

Emacs may not build using some Cygwin builds of GCC, such as Cygwin
version 1.1.8, using the default configure settings.  It appears to be
necessary to specify the -mwin32 flag when compiling, and define
__MSVCRT__, like so:

  configure --with-gcc --cflags -mwin32 --cflags -D__MSVCRT__

*** Building the MS-Windows port fails with a CreateProcess failure.

Some versions of mingw32 make on some versions of Windows do not seem
to detect the shell correctly. Try "make SHELL=cmd.exe", or if that
fails, try running make from Cygwin bash instead.

*** Building `ctags' for MS-Windows with the MinGW port of GCC fails.

This might happen due to a bug in the MinGW header assert.h, which
defines the `assert' macro with a trailing semi-colon.  The following
patch to assert.h should solve this:

 *** include/assert.h.orig      Sun Nov  7 02:41:36 1999
 --- include/assert.h   Mon Jan 29 11:49:10 2001
 ***************
 *** 41,47 ****
   /*
    * If not debugging, assert does nothing.
    */
 ! #define assert(x)    ((void)0);

   #else /* debugging enabled */

 --- 41,47 ----
   /*
    * If not debugging, assert does nothing.
    */
 ! #define assert(x)    ((void)0)

   #else /* debugging enabled */


*** Building the MS-Windows port with Visual Studio 2005 fails.

Microsoft no longer ships the single threaded version of the C library
with their compiler, and the multithreaded static library is missing
some functions that Microsoft have deemed non-threadsafe.  The
dynamically linked C library has all the functions, but there is a
conflict between the versions of malloc in the DLL and in Emacs, which
is not resolvable due to the way Windows does dynamic linking.

We recommend the use of the MinGW port of GCC for compiling Emacs, as
not only does it not suffer these problems, but it is also Free
software like Emacs.

*** Building the MS-Windows port with Visual Studio fails compiling emacs.rc

If the build fails with the following message then the problem
described here most likely applies:

../nt/emacs.rc(1) : error RC2176 : old DIB in icons\emacs.ico; pass it
through SDKPAINT

The Emacs icon contains a high resolution PNG icon for Vista, which is
not recognized by older versions of the resource compiler. There are
several workarounds for this problem:
        1. Use Free MinGW tools to compile, which do not have this problem.
        2. Install the latest Windows SDK.
        3. Replace emacs.ico with an older or edited icon.

*** Building the MS-Windows port complains about unknown escape sequences.

Errors and warnings can look like this:

 w32.c:1959:27: error: \x used with no following hex digits
 w32.c:1959:27: warning: unknown escape sequence '\i'

This happens when paths using backslashes are passed to the compiler or
linker (via -I and possibly other compiler flags); when these paths are
included in source code, the backslashes are interpreted as escape sequences.
See http://lists.gnu.org/archive/html/emacs-devel/2010-07/msg00995.html

The fix is to use forward slashes in all paths passed to the compiler.

** Linking

*** Building Emacs with a system compiler fails to link because of an
undefined symbol such as __eprintf which does not appear in Emacs.

This can happen if some of the libraries linked into Emacs were built
with GCC, but Emacs itself is being linked with a compiler other than
GCC.  Object files compiled with GCC might need some helper functions
from libgcc.a, the library which comes with GCC, but the system
compiler does not instruct the linker to search libgcc.a during the
link stage.

A solution is to link with GCC, like this:

        make CC=gcc

Since the .o object files already exist, this will not recompile Emacs
with GCC, but just restart by trying again to link temacs.

*** Sun with acc: Link failure when using acc on a Sun.

To use acc, you need additional options just before the libraries, such as

   /usr/lang/SC2.0.1/values-Xt.o -L/usr/lang/SC2.0.1/cg87 -L/usr/lang/SC2.0.1

and you need to add -lansi just before -lc.

The precise file names depend on the compiler version, so we
cannot easily arrange to supply them.

*** Linking says that the functions insque and remque are undefined.

Change oldXMenu/Makefile by adding insque.o to the variable OBJS.

*** `tparam' reported as a multiply-defined symbol when linking with ncurses.

This problem results from an incompatible change in ncurses, in
version 1.9.9e approximately.  This version is unable to provide a
definition of tparm without also defining tparam.  This is also
incompatible with Terminfo; as a result, the Emacs Terminfo support
does not work with this version of ncurses.

The fix is to install a newer version of ncurses, such as version 4.2.

** Bootstrapping

Bootstrapping (compiling the .el files) is normally only necessary
with development builds, since the .elc files are pre-compiled in releases.

*** "No rule to make target" with Ubuntu 8.04 make 3.81-3build1

Compiling the lisp files fails at random places, complaining:
"No rule to make target `/path/to/some/lisp.elc'".
The causes of this problem are not understood.  Using GNU make 3.81 compiled
from source, rather than the Ubuntu version, worked. See Bug#327,821.

** Dumping

*** Linux: Segfault during `make bootstrap' under certain recent versions of the Linux kernel.

With certain recent Linux kernels (like the one of Red Hat Fedora Core
1 and newer), the new "Exec-shield" functionality is enabled by default, which
creates a different memory layout that breaks the emacs dumper.  Emacs tries
to handle this at build time, but if the workaround used fails, these
instructions can be useful.
The work-around explained here is not enough on Fedora Core 4 (and possible
newer). Read the next item.

Configure can overcome the problem of exec-shield if the architecture is
x86 and the program setarch is present.  On other architectures no
workaround is known.

You can check the Exec-shield state like this:

    cat /proc/sys/kernel/exec-shield

It returns non-zero when Exec-shield is enabled, 0 otherwise.  Please
read your system documentation for more details on Exec-shield and
associated commands.  Exec-shield can be turned off with this command:

    echo "0" > /proc/sys/kernel/exec-shield

When Exec-shield is enabled, building Emacs will segfault during the
execution of this command:

    ./temacs --batch --load loadup [dump|bootstrap]

To work around this problem, it is necessary to temporarily disable
Exec-shield while building Emacs, or, on x86, by using the `setarch'
command when running temacs like this:

    setarch i386 ./temacs --batch --load loadup [dump|bootstrap]


*** Fedora Core 4 GNU/Linux: Segfault during dumping.

In addition to exec-shield explained above "Linux: Segfault during
`make bootstrap' under certain recent versions of the Linux kernel"
item, Linux kernel shipped with Fedora Core 4 randomizes the virtual
address space of a process. As the result dumping may fail even if
you turn off exec-shield. In this case, use the -R option to the setarch
command:

   setarch i386 -R ./temacs --batch --load loadup [dump|bootstrap]

or

   setarch i386 -R make bootstrap

*** Fatal signal in the command  temacs -l loadup inc dump.

This command is the final stage of building Emacs.  It is run by the
Makefile in the src subdirectory.

It has been known to get fatal errors due to insufficient swapping
space available on the machine.

On 68000s, it has also happened because of bugs in the
subroutine `alloca'.  Verify that `alloca' works right, even
for large blocks (many pages).

*** test-distrib says that the distribution has been clobbered.
*** or, temacs prints "Command key out of range 0-127".
*** or, temacs runs and dumps emacs, but emacs totally fails to work.
*** or, temacs gets errors dumping emacs.

This can be because the .elc files have been garbled.  Do not be
fooled by the fact that most of a .elc file is text: these are
binary files and can contain all 256 byte values.

In particular `shar' cannot be used for transmitting GNU Emacs.
It typically truncates "lines".  What appear to be "lines" in
a binary file can of course be of any length.  Even once `shar'
itself is made to work correctly, `sh' discards null characters
when unpacking the shell archive.

I have also seen character \177 changed into \377.  I do not know
what transfer means caused this problem.  Various network
file transfer programs are suspected of clobbering the high bit.

If you have a copy of Emacs that has been damaged in its
nonprinting characters, you can fix them:

 1) Record the names of all the .elc files.
 2) Delete all the .elc files.
 3) Recompile alloc.c with a value of PURESIZE twice as large.
     (See puresize.h.)  You might as well save the old alloc.o.
 4) Remake emacs.  It should work now.
 5) Running emacs, do Meta-x byte-compile-file repeatedly
  to recreate all the .elc files that used to exist.
  You may need to increase the value of the variable
  max-lisp-eval-depth to succeed in running the compiler interpreted
  on certain .el files.  400 was sufficient as of last report.
 6) Reinstall the old alloc.o (undoing changes to alloc.c if any)
  and remake temacs.
 7) Remake emacs.  It should work now, with valid .elc files.

*** temacs prints "Pure Lisp storage exhausted".

This means that the Lisp code loaded from the .elc and .el files
during temacs -l loadup inc dump took up more space than was allocated.

This could be caused by
 1) adding code to the preloaded Lisp files
 2) adding more preloaded files in loadup.el
 3) having a site-init.el or site-load.el which loads files.
   Note that ANY site-init.el or site-load.el is nonstandard;
   if you have received Emacs from some other site and it contains a
   site-init.el or site-load.el file, consider deleting that file.
 4) getting the wrong .el or .elc files
   (not from the directory you expected).
 5) deleting some .elc files that are supposed to exist.
   This would cause the source files (.el files) to be
   loaded instead.  They take up more room, so you lose.
 6) a bug in the Emacs distribution which underestimates the space required.

If the need for more space is legitimate, change the definition
of PURESIZE in puresize.h.

But in some of the cases listed above, this problem is a consequence
of something else that is wrong.  Be sure to check and fix the real problem.

*** Linux: Emacs crashes when dumping itself on Mac PPC running Yellow Dog GNU/Linux.

The crashes happen inside the function Fmake_symbol; here's a typical
C backtrace printed by GDB:

  0x190c0c0 in Fmake_symbol ()
  (gdb) where
  #0  0x190c0c0 in Fmake_symbol ()
  #1  0x1942ca4 in init_obarray ()
  #2  0x18b3500 in main ()
  #3  0x114371c in __libc_start_main (argc=5, argv=0x7ffff5b4, envp=0x7ffff5cc,

This could happen because GCC version 2.95 and later changed the base
of the load address to 0x10000000.  Emacs needs to be told about this,
but we currently cannot do that automatically, because that breaks
other versions of GNU/Linux on the MacPPC.  Until we find a way to
distinguish between the Yellow Dog and the other varieties of
GNU/Linux systems on the PPC, you will have to manually uncomment the
following section near the end of the file src/m/macppc.h in the Emacs
distribution:

  #if 0  /* This breaks things on PPC GNU/Linux except for Yellowdog,
            even with identical GCC, as, ld.  Let's take it out until we
            know what's really going on here.  */
  /* GCC 2.95 and newer on GNU/Linux PPC changed the load address to
     0x10000000.  */
  #if defined __linux__
  #if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)
  #define DATA_SEG_BITS  0x10000000
  #endif
  #endif
  #endif /* 0 */

Remove the "#if 0" and "#endif" directives which surround this, save
the file, and then reconfigure and rebuild Emacs.  The dumping process
should now succeed.

*** OpenBSD 4.0 macppc: Segfault during dumping.

The build aborts with signal 11 when the command `./temacs --batch
--load loadup bootstrap' tries to load files.el.  A workaround seems
to be to reduce the level of compiler optimization used during the
build (from -O2 to -O1).  It is possible this is an OpenBSD
GCC problem specific to the macppc architecture, possibly only
occurring with older versions of GCC (e.g. 3.3.5).

*** openSUSE 10.3: Segfault in bcopy during dumping.

This is due to a bug in the bcopy implementation in openSUSE 10.3.
It is/will be fixed in an openSUSE update.

** Installation

*** Installing Emacs gets an error running `install-info'.

You need to install a recent version of Texinfo; that package
supplies the `install-info' command.

*** Installing to a directory with spaces in the name fails.

For example, if you call configure with a directory-related option
with spaces in the value, eg --enable-locallisppath='/path/with\ spaces'.
Using directory paths with spaces is not supported at this time: you
must re-configure without using spaces.

*** Installing to a directory with non-ASCII characters in the name fails.

Installation may fail, or the Emacs executable may not start
correctly, if a directory name containing non-ASCII characters is used
as a `configure' argument (e.g. `--prefix').  The problem can also
occur if a non-ASCII directory is specified in the EMACSLOADPATH
envvar.

*** On Solaris, use GNU Make when installing an out-of-tree build

The Emacs configuration process allows you to configure the
build environment so that you can build emacs in a directory
outside of the distribution tree.  When installing Emacs from an
out-of-tree build directory on Solaris, you may need to use GNU
make.  The make programs bundled with Solaris support the VPATH
macro but use it differently from the way the VPATH macro is
used by GNU make.  The differences will cause the "make install"
step to fail, leaving you with an incomplete emacs
installation.  GNU make is available in /usr/sfw/bin on Solaris
10 and can be installed as /opt/sfw/bin/gmake from the Solaris 9
Software Companion CDROM.

The problems due to the VPATH processing differences affect only
out of tree builds so, if you are on a Solaris installation
without GNU make, you can install Emacs completely by installing
from a build environment using the original emacs distribution tree.

** First execution

*** Emacs binary is not in executable format, and cannot be run.

This was reported to happen when Emacs is built in a directory mounted
via NFS, for some combinations of NFS client and NFS server.
Usually, the file `emacs' produced in these cases is full of
binary null characters, and the `file' utility says:

    emacs: ASCII text, with no line terminators

We don't know what exactly causes this failure.  A work-around is to
build Emacs in a directory on a local disk.

*** The dumped Emacs crashes when run, trying to write pure data.

Two causes have been seen for such problems.

1) On a system where getpagesize is not a system call, it is defined
as a macro.  If the definition (in both unex*.c and malloc.c) is wrong,
it can cause problems like this.  You might be able to find the correct
value in the man page for a.out (5).

2) Some systems allocate variables declared static among the
initialized variables.  Emacs makes all initialized variables in most
of its files pure after dumping, but the variables declared static and
not initialized are not supposed to be pure.  On these systems you
may need to add "#define static" to the m- or the s- file.

* Runtime problems on legacy systems

This section covers bugs reported on very old hardware or software.
If you are using hardware and an operating system shipped after 2000,
it is unlikely you will see any of these.

*** OPENSTEP 4.2: Compiling syntax.c with gcc 2.7.2.1 fails.

The compiler was reported to crash while compiling syntax.c with the
following message:

   cc: Internal compiler error: program cc1obj got fatal signal 11

To work around this, replace the macros UPDATE_SYNTAX_TABLE_FORWARD,
INC_BOTH, and INC_FROM with functions.  To this end, first define 3
functions, one each for every macro.  Here's an example:

    static int update_syntax_table_forward(int from)
    {
        return(UPDATE_SYNTAX_TABLE_FORWARD(from));
    }/*update_syntax_table_forward*/

Then replace all references to UPDATE_SYNTAX_TABLE_FORWARD in syntax.c
with a call to the function update_syntax_table_forward.

*** Solaris 2.x

**** Strange results from format %d in a few cases, on a Sun.

Sun compiler version SC3.0 has been found to miscompile part of
editfns.c.  The workaround is to compile with some other compiler such
as GCC.

**** On Solaris, Emacs dumps core if lisp-complete-symbol is called.

If you compile Emacs with the -fast or -xO4 option with version 3.0.2
of the Sun C compiler, Emacs dumps core when lisp-complete-symbol is
called.  The problem does not happen if you compile with GCC.

**** On Solaris, Emacs crashes if you use (display-time).

This can happen if you configure Emacs without specifying the precise
version of Solaris that you are using.

**** Solaris 2.x: GCC complains "64 bit integer types not supported".

This suggests that GCC is not installed correctly.  Most likely you
are using GCC 2.7.2.3 (or earlier) on Solaris 2.6 (or later); this
does not work without patching.  To run GCC 2.7.2.3 on Solaris 2.6 or
later, you must patch fixinc.svr4 and reinstall GCC from scratch as
described in the Solaris FAQ
<http://www.wins.uva.nl/pub/solaris/solaris2.html>.  A better fix is
to upgrade to GCC 2.8.1 or later.

**** Solaris 2.7: Building Emacs with WorkShop Compilers 5.0 98/12/15
C 5.0 failed, apparently with non-default CFLAGS, most probably due to
compiler bugs.  Using Sun Solaris 2.7 Sun WorkShop 6 update 1 C
release was reported to work without problems.  It worked OK on
another system with Solaris 8 using apparently the same 5.0 compiler
and the default CFLAGS.

**** Solaris 2.x: Emacs dumps core when built with Motif.

The Solaris Motif libraries are buggy, at least up through Solaris 2.5.1.
Install the current Motif runtime library patch appropriate for your host.
(Make sure the patch is current; some older patch versions still have the bug.)
You should install the other patches recommended by Sun for your host, too.
You can obtain Sun patches from ftp://sunsolve.sun.com/pub/patches/;
look for files with names ending in `.PatchReport' to see which patches
are currently recommended for your host.

On Solaris 2.6, Emacs is said to work with Motif when Solaris patch
105284-12 is installed, but fail when 105284-15 is installed.
105284-18 might fix it again.

**** Solaris 2.6 and 7: the Compose key does not work.

This is a bug in Motif in Solaris.  Supposedly it has been fixed for
the next major release of Solaris.  However, if someone with Sun
support complains to Sun about the bug, they may release a patch.
If you do this, mention Sun bug #4188711.

One workaround is to use a locale that allows non-ASCII characters.
For example, before invoking emacs, set the LC_ALL environment
variable to "en_US" (American English).  The directory /usr/lib/locale
lists the supported locales; any locale other than "C" or "POSIX"
should do.

pen@lysator.liu.se says (Feb 1998) that the Compose key does work
if you link with the MIT X11 libraries instead of the Solaris X11 libraries.

*** HP/UX 10: Large file support is disabled.
(HP/UX 10 was end-of-lifed in May 1999.)
See the comments in src/s/hpux10-20.h.

*** HP/UX: Emacs is slow using X11R5.

This happens if you use the MIT versions of the X libraries--it
doesn't run as fast as HP's version.  People sometimes use the version
because they see the HP version doesn't have the libraries libXaw.a,
libXmu.a, libXext.a and others.  HP/UX normally doesn't come with
those libraries installed.  To get good performance, you need to
install them and rebuild Emacs.

*** UnixWare 2.1: Error 12 (virtual memory exceeded) when dumping Emacs.

Paul Abrahams (abrahams@acm.org) reports that with the installed
virtual memory settings for UnixWare 2.1.2, an Error 12 occurs during
the "make" that builds Emacs, when running temacs to dump emacs.  That
error indicates that the per-process virtual memory limit has been
exceeded.  The default limit is probably 32MB.  Raising the virtual
memory limit to 40MB should make it possible to finish building Emacs.

You can do this with the command `ulimit' (sh) or `limit' (csh).
But you have to be root to do it.

According to Martin Sohnius, you can also retune this in the kernel:

    # /etc/conf/bin/idtune SDATLIM 33554432         ## soft data size limit
    # /etc/conf/bin/idtune HDATLIM 33554432         ## hard "
    # /etc/conf/bin/idtune SVMMSIZE unlimited       ## soft process size limit
    # /etc/conf/bin/idtune HVMMSIZE unlimited       ## hard "
    # /etc/conf/bin/idbuild -B

(He recommends you not change the stack limit, though.)
These changes take effect when you reboot.

** MS-Windows 95, 98, ME, and NT

*** MS-Windows NT/95: Problems running Perl under Emacs

`perl -de 0' just hangs when executed in an Emacs subshell.
The fault lies with Perl (indirectly with Windows NT/95).

The problem is that the Perl debugger explicitly opens a connection to
"CON", which is the DOS/NT equivalent of "/dev/tty", for interacting
with the user.

On Unix, this is okay, because Emacs (or the shell?) creates a
pseudo-tty so that /dev/tty is really the pipe Emacs is using to
communicate with the subprocess.

On NT, this fails because CON always refers to the handle for the
relevant console (approximately equivalent to a tty), and cannot be
redirected to refer to the pipe Emacs assigned to the subprocess as
stdin.

A workaround is to modify perldb.pl to use STDIN/STDOUT instead of CON.

For Perl 4:

    *** PERL/LIB/PERLDB.PL.orig Wed May 26 08:24:18 1993
    --- PERL/LIB/PERLDB.PL      Mon Jul 01 15:28:16 1996
    ***************
    *** 68,74 ****
          $rcfile=".perldb";
      }
      else {
    !     $console = "con";
          $rcfile="perldb.ini";
      }

    --- 68,74 ----
          $rcfile=".perldb";
      }
      else {
    !     $console = "";
          $rcfile="perldb.ini";
      }


    For Perl 5:
    *** perl/5.001/lib/perl5db.pl.orig  Sun Jun 04 21:13:40 1995
    --- perl/5.001/lib/perl5db.pl       Mon Jul 01 17:00:08 1996
    ***************
    *** 22,28 ****
          $rcfile=".perldb";
      }
      elsif (-e "con") {
    !     $console = "con";
          $rcfile="perldb.ini";
      }
      else {
    --- 22,28 ----
          $rcfile=".perldb";
      }
      elsif (-e "con") {
    !     $console = "";
          $rcfile="perldb.ini";
      }
      else {

*** MS-Windows 95: Alt-f6 does not get through to Emacs.

This character seems to be trapped by the kernel in Windows 95.
You can enter M-f6 by typing ESC f6.

*** MS-Windows 95/98/ME: subprocesses do not terminate properly.

This is a limitation of the Operating System, and can cause problems
when shutting down Windows. Ensure that all subprocesses are exited
cleanly before exiting Emacs. For more details, see the FAQ at
http://www.gnu.org/software/emacs/windows/.

*** MS-Windows 95/98/ME: crashes when Emacs invokes non-existent programs.

When a program you are trying to run is not found on the PATH,
Windows might respond by crashing or locking up your system.  In
particular, this has been reported when trying to compile a Java
program in JDEE when javac.exe is installed, but not on the system PATH.

** MS-DOS

*** When compiling with DJGPP on MS-Windows NT or later, "config msdos" fails.

If the error message is "VDM has been already loaded", this is because
Windows has a program called `redir.exe' that is incompatible with a
program by the same name supplied with DJGPP, which is used by
config.bat.  To resolve this, move the DJGPP's `bin' subdirectory to
the front of your PATH environment variable.

*** When Emacs compiled with DJGPP runs on Windows 2000 and later, it cannot
find your HOME directory.

This was reported to happen when you click on "Save for future
sessions" button in a Customize buffer.  You might see an error
message like this one:

  basic-save-buffer-2: c:/FOO/BAR/~dosuser/: no such directory

(The telltale sign is the "~USER" part at the end of the directory
Emacs complains about, where USER is your username or the literal
string "dosuser", which is the default username set up by the DJGPP
startup file DJGPP.ENV.)

This happens when the functions `user-login-name' and
`user-real-login-name' return different strings for your username as
Emacs sees it.  To correct this, make sure both USER and USERNAME
environment variables are set to the same value.  Windows 2000 and
later sets USERNAME, so if you want to keep that, make sure USER is
set to the same value.  If you don't want to set USER globally, you
can do it in the [emacs] section of your DJGPP.ENV file.

*** When Emacs compiled with DJGPP runs on Vista, it runs out of memory.

If Emacs running on Vista displays "!MEM FULL!" in the mode line, you
are hitting the memory allocation bugs in the Vista DPMI server.  See
msdos/INSTALL for how to work around these bugs (search for "Vista").

*** When compiling with DJGPP on MS-Windows 95, Make fails for some targets
like make-docfile.

This can happen if long file name support (the setting of environment
variable LFN) when Emacs distribution was unpacked and during
compilation are not the same.  See msdos/INSTALL for the explanation
of how to avoid this problem.

*** Emacs compiled with DJGPP complains at startup:

  "Wrong type of argument: internal-facep, msdos-menu-active-face"

This can happen if you define an environment variable `TERM'.  Emacs
on MSDOS uses an internal terminal emulator which is disabled if the
value of `TERM' is anything but the string "internal".  Emacs then
works as if its terminal were a dumb glass teletype that doesn't
support faces.  To work around this, arrange for `TERM' to be
undefined when Emacs runs.  The best way to do that is to add an
[emacs] section to the DJGPP.ENV file which defines an empty value for
`TERM'; this way, only Emacs gets the empty value, while the rest of
your system works as before.

*** MS-DOS: Emacs crashes at startup.

Some users report that Emacs 19.29 requires dpmi memory management,
and crashes on startup if the system does not have it.  We don't
know why this happens--perhaps these machines don't have enough real
memory, or perhaps something is wrong in Emacs or the compiler.
However, arranging to use dpmi support is a workaround.

You can find out if you have a dpmi host by running go32 without
arguments; it will tell you if it uses dpmi memory.  For more
information about dpmi memory, consult the djgpp FAQ.  (djgpp
is the GNU C compiler as packaged for MSDOS.)

Compiling Emacs under MSDOS is extremely sensitive for proper memory
configuration.  If you experience problems during compilation, consider
removing some or all memory resident programs (notably disk caches)
and make sure that your memory managers are properly configured.  See
the djgpp faq for configuration hints.

*** Emacs compiled with DJGPP for MS-DOS/MS-Windows cannot access files
in the directory with the special name `dev' under the root of any
drive, e.g. `c:/dev'.

This is an unfortunate side-effect of the support for Unix-style
device names such as /dev/null in the DJGPP runtime library.  A
work-around is to rename the problem directory to another name.

*** MS-DOS+DJGPP: Problems on MS-DOS if DJGPP v2.0 is used to compile Emacs.

There are two DJGPP library bugs which cause problems:

  * Running `shell-command' (or `compile', or `grep') you get
    `Searching for program: permission denied (EACCES), c:/command.com';
  * After you shell to DOS, Ctrl-Break kills Emacs.

To work around these bugs, you can use two files in the msdos
subdirectory: `is_exec.c' and `sigaction.c'.  Compile them and link
them into the Emacs executable `temacs'; then they will replace the
incorrect library functions.

*** MS-DOS: Emacs compiled for MSDOS cannot find some Lisp files, or other
run-time support files, when long filename support is enabled.

Usually, this problem will manifest itself when Emacs exits
immediately after flashing the startup screen, because it cannot find
the Lisp files it needs to load at startup.  Redirect Emacs stdout
and stderr to a file to see the error message printed by Emacs.

Another manifestation of this problem is that Emacs is unable to load
the support for editing program sources in languages such as C and Lisp.

This can happen if the Emacs distribution was unzipped without LFN
support, thus causing long filenames to be truncated to the first 6
characters and a numeric tail that Windows 95 normally attaches to it.
You should unzip the files again with a utility that supports long
filenames (such as djtar from DJGPP or InfoZip's UnZip program
compiled with DJGPP v2).  The file msdos/INSTALL explains this issue
in more detail.

Another possible reason for such failures is that Emacs compiled for
MSDOS is used on Windows NT, where long file names are not supported
by this version of Emacs, but the distribution was unpacked by an
unzip program that preserved the long file names instead of truncating
them to DOS 8+3 limits.  To be useful on NT, the MSDOS port of Emacs
must be unzipped by a DOS utility, so that long file names are
properly truncated.

** Archaic window managers and toolkits

*** OpenLook: Under OpenLook, the Emacs window disappears when you type M-q.

Some versions of the Open Look window manager interpret M-q as a quit
command for whatever window you are typing at.  If you want to use
Emacs with that window manager, you should try to configure the window
manager to use some other command.   You can disable the
shortcut keys entirely by adding this line to ~/.OWdefaults:

    OpenWindows.WindowMenuAccelerators: False

*** twm: A position you specified in .Xdefaults is ignored, using twm.

twm normally ignores "program-specified" positions.
You can tell it to obey them with this command in your `.twmrc' file:

  UsePPosition  "on"            #allow clients to request a position

** Bugs related to old DEC hardware

*** The Compose key on a DEC keyboard does not work as Meta key.

This shell command should fix it:

  xmodmap -e 'keycode 0xb1 = Meta_L'

*** Keyboard input gets confused after a beep when using a DECserver
as a concentrator.

This problem seems to be a matter of configuring the DECserver to use
7 bit characters rather than 8 bit characters.

* Build problems on legacy systems

** SunOS: Emacs gets error message from linker on Sun.

If the error message says that a symbol such as `f68881_used' or
`ffpa_used' or `start_float' is undefined, this probably indicates
that you have compiled some libraries, such as the X libraries,
with a floating point option other than the default.

It's not terribly hard to make this work with small changes in
crt0.c together with linking with Fcrt1.o, Wcrt1.o or Mcrt1.o.
However, the easiest approach is to build Xlib with the default
floating point option: -fsoft.

** HPUX 10.20: Emacs crashes during dumping on the HPPA machine.

This seems to be due to a GCC bug; it is fixed in GCC 2.8.1.

** Vax C compiler bugs affecting Emacs.

You may get one of these problems compiling Emacs:

   foo.c line nnn: compiler error: no table entry for op STASG
   foo.c: fatal error in /lib/ccom

These are due to bugs in the C compiler; the code is valid C.
Unfortunately, the bugs are unpredictable: the same construct
may compile properly or trigger one of these bugs, depending
on what else is in the source file being compiled.  Even changes
in header files that should not affect the file being compiled
can affect whether the bug happens.  In addition, sometimes files
that compile correctly on one machine get this bug on another machine.

As a result, it is hard for me to make sure this bug will not affect
you.  I have attempted to find and alter these constructs, but more
can always appear.  However, I can tell you how to deal with it if it
should happen.  The bug comes from having an indexed reference to an
array of Lisp_Objects, as an argument in a function call:
  Lisp_Object *args;
  ...
   ... foo (5, args[i], ...)...
putting the argument into a temporary variable first, as in
  Lisp_Object *args;
  Lisp_Object tem;
  ...
   tem = args[i];
   ... foo (r, tem, ...)...
causes the problem to go away.
The `contents' field of a Lisp vector is an array of Lisp_Objects,
so you may see the problem happening with indexed references to that.

** 68000 C compiler problems

Various 68000 compilers have different problems.
These are some that have been observed.

*** Using value of assignment expression on union type loses.
This means that  x = y = z;  or  foo (x = z);  does not work
if x is of type Lisp_Object.

*** "cannot reclaim" error.

This means that an expression is too complicated.  You get the correct
line number in the error message.  The code must be rewritten with
simpler expressions.

*** XCONS, XSTRING, etc macros produce incorrect code.

If temacs fails to run at all, this may be the cause.
Compile this test program and look at the assembler code:

struct foo { char x; unsigned int y : 24; };

lose (arg)
     struct foo arg;
{
  test ((int *) arg.y);
}

If the code is incorrect, your compiler has this problem.
In the XCONS, etc., macros in lisp.h you must replace (a).u.val with
((a).u.val + coercedummy) where coercedummy is declared as int.

This problem will only happen if USE_LISP_UNION_TYPE is manually
defined in lisp.h.

** C compilers lose on returning unions.

I hear that some C compilers cannot handle returning a union type.
Most of the functions in GNU Emacs return type Lisp_Object, which is
defined as a union on some rare architectures.

This problem will only happen if USE_LISP_UNION_TYPE is manually
defined in lisp.h.

\f
This file is part of GNU Emacs.

GNU Emacs is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

GNU Emacs is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with GNU Emacs.  If not, see <http://www.gnu.org/licenses/>.

\f
Local variables:
mode: outline
paragraph-separate: "[  \f]*$"
end: