*** empty log message ***

[arla.git] / ChangeLog.2002

2002-12-30  Love  <lha@stacken.kth.se>

        * appl/vos/vos.8: fix Dt section
        
        * appl/bos/bos.8: fix Dt section
        
        * tests: switch to emacs-21.2

        * tests/generic-build: send error output to fd 4 (progress)
        
2002-12-29  Love  <lha@stacken.kth.se>

        * configure.in (bsd): don't check for kernel make_dev, we no
        longer need it

        * nnpfs/bsd/nnpfs_wrap-bsd.c (make_devices/destroy_devices):
        remove ifdef on HAVE_KERNEL_MAKE_DEV since we no longer support
        pre freebsd 4.1, rename remove_devices to destroy_devices to match
        destroy_dev, also remove support for CDEV_MODULE (freebsd 4.0)
        
        * appl/mpp/mpp.c (mpp_isblank): returns an int
        
        * nnpfs/bsd/Makefile.in: fix pre and post ld magic helper objects
        for freebsd < 5

        * conf/Makefile.in: (clean): remove arla.spec
        
        * configure.in: (AC_OUTPUT): don't build conf/arla.conf and
        conf/arla.spec, let the makefile do that.
        
2002-12-21  Love  <lha@stacken.kth.se>

        * rx/rxperf.c: s/strings.h/string.h/ for mem* functions

        * scripts/arla-release.sh: autom4te.cache is a directory

        * configure.in: pre7

        * HACKING: don't fail if automake failes for now

        * appl/fs/fs.1: s/comand/command/; from openbsd

2002-12-20  Love  <lha@stacken.kth.se>

        * arlad/fcache.c (fcache_init): don't set fprioritylevel, its
        already set
        
        * INSTALL: update to match configure options
        
        * INSTALL: no mmaptime
        
        * configure.in: no mmaptime
        
        * util/mmaptime.h: OBSOLETE
        
        * util/mmaptime.c: OBSOLETE
        
        * lwp/fasttime.c: no mmaptime
        
        * rxkad/rxkad_locl.h: no mmaptime
        
        * rx/rx_locl.h: no mmaptime
        
        * arlad/arla_local.h: no mmaptime
        
        * nnpfs/bsd/bin/startarla.in: move mknod after kldload so devfs
        have its chance to do its magic dance

        * doc/real-world.texi: update port and natkeep stuff
        
        * INSTALL: update port info
        
        * configure.in: if netbsd, set default_sys to /usr/src/sys
        
2002-12-19  Love  <lha@stacken.kth.se>

        * configure.in: pre6

        * arlad/fcache.[ch]: split find_first_fs into two parts,
        init_fs_context and find_first_fs.  now the vldb operation can
        return error values to caller. This fixes problem haba had with
        expired token in root's pag

        * conf/arla.conf.in: remove fpriority as a configurable option, to
        be folded into the connection level instead
        
        * arlad/arla.{c,conf.in}: remove fpriority as a configurable
        option, to be folded into the connection level instead

        * configure.in: remove test for mnt_nvnodelist

        * nnpfs/bsd: change how nnpfs nodes are looked up to a hash
        instead of a linked list (the struct mount->m_vnodes list). This
        removes a race, an autoconf test and make it faster

        * conf/arla.conf{,.in}: generate arla.conf so we can changes the
        value depending on target platform

        * configure.in: ARLA_CONF_*: move arla.conf tunable option to here
        
        * tests/run-tests.in: make date in test output less verbose
        
2002-12-18  Love  <lha@stacken.kth.se>

        * NEWS: FreeBSD 5.0 is in 0.36
        
        * scripts/arla-release.sh: remove autom4te.cache if there is a
        $HOME/.gnupg, sign tar-boll
        
        * nnpfs/bsd/nnpfs_common-bsd.c (MALLOC_DEFINE): make helptests
        better
        
        * configure.in: pre5
        
        * nnpfs/bsd/nnpfs_syscalls-wrap-freebsd.c: force install afs
        syscall

        * tests/build-gdb: 5.3 is released
        
        * nnpfs/bsd/nnpfs_vnodeops-bsd.c: add debug output to
        nnpfs_pathconf

        * nnpfs/bsd/nnpfs_vnodeops-bsd.c: add vop_pathconf
        
        * configure.in: check for vop_stdpathconf
        
        * nnpfs/bsd/nnpfs_vnodeops-common.c (nnpfs_read_common): make sure
        we don't touch the vnode after we have done a vrecycle(), also
        make sure we don't look at the content of the vnode after we have
        release the vnode

        * nnpfs/bsd: split the malloc type in 4 diffrent malloc types

        * nnpfs/bsd/nnpfs_vnodeops-bsd.c (nnpfs_link): FreeBSD 5.0 doesn't
        need to lock the vnode in VOP_LINK

        * nnpfs/bsd/nnpfs/nnpfs_locl.h: move NNPFS_MAKE_VROOT to
        nnpfs_locl.h include <sys/vnode.h>
        
        * nnpfs/bsd/nnpfs_vnodeops-bsd.c (nnpfs_abortop,cleanup_cnp):
        remember to clear the HASBUF flag when when freeing the component
        name in the uma_zfree case

2002-12-17  Love  <lha@stacken.kth.se>

        * configure.in: add some freebsd 5 CFLAGS
        
        * nnpfs/bsd/nnpfs_vnodeops-common.c: comment out cruft after
        #endif/#else

        * nnpfs/bsd/nnpfs/nnpfs_locl.h: if exists, include <vm/uma.h>
        prototype for (freebsd5) compat VT_ defines
        
        * nnpfs/bsd/nnpfs_vfsops-freebsd.c: provide compat VT_ defines
        
        * nnpfs/bsd/nnpfs_message.c: use NNPFS_MAKE_VROOT comment out text
        after #endif
        
        * nnpfs/bsd/nnpfs/nnpfs_common.h: add helper macro
        NNPFS_MAKE_VROOT that sets that VROOT flag on a vnode this is
        since freebsd 5 split that v_flag field into too parts v_iflag and
        v_vflag because of locking issues

        * tests/run-tests.in: make sure the hostname is always set to
        something so we don't pass a - to mkdir
        
        * nnpfs/bsd/nnpfs_dev-common.c (nnpfs_cursig): its not easy to
        find the CURSIG for freebsd5, lets just ignore it for now (and
        return 0)
        
        * nnpfs/bsd/Makefile.in: build glue for freebsd5
        
        * configure.in: add some freebsd 5 glue: add automake variable
        FBSD5, check for <vm/uma.h>, and uma_zfree_arg()
        
        * nnpfs/bsd/Makefile.in: yet another way to build vnode_if.h
        (freebsd 5.0)

2002-12-16  Love  <lha@stacken.kth.se>

        * arlad/arla.conf.5: s/maxium/maximum/g ; Jesper Louis Andersen;
        from OpenBSD

2002-12-15  Love  <lha@stacken.kth.se>

        * rx/rx.c: remove more rx_{Read,Write}Proc
        
        * rx/rx_user.h: cleanup, ident, don't provide a prototype for
        fprintf

        * rx/rx_user.h: kill USE_MMAPTIME
        
        * rx/rx_rdwr.c: remove all rx_{Read,Write}proc and use
        rx_Write/rx_Read

        * rx/rx.h: ident some functions remove all rx_{Read,Write}proc and
        use rx_Write/rx_Read
        
        * arlad/messages.c (possibly_have_network): Ignore `0' as an ipv4
        address it wont work either. Comment on what v6 addresses to
        ignore.

2002-12-11  Love  <lha@stacken.kth.se>

        * rxkad/rxk_crpt.c: remove rx_SlowPutInt32 proto, no longer used
        
        * arlad/messages.c (viocconnect): when reconnecting, get a new
        uuid to make sure the fileserver doesn't get upset with us
        
2002-12-07  Love  <lha@stacken.kth.se>

        * configure.in: remember the flags we tested comerr with
        
2002-12-06  Love  <lha@stacken.kth.se>

        * util/log.c: avoid using reserved word 'log'
        
        * lib/ko/kocell.c: avoid using reserved word 'log'
        
        * arlad/arladeb.h: avoid using reserved word 'log'
        
        * arlad/arladeb.c: avoid using reserved word 'log'
        
2002-12-04  Tomas Olsson  <tol@stacken.kth.se>

        * nnpfs/bsd/nnpfs_syscalls-wrap-bsd.c (try_install_syscall): on
        openbsd, ignore that AFS_SYSCALL appears used

2002-12-03  Love  <lha@stacken.kth.se>

        * appl/mac/install/StartupParameters.plist: add missing ;

        * appl/mac/make-mac-package.in: cleanup
        
        * NEWS: Changes in release 0.35.11
        
2002-12-03  Tomas Olsson  <tol@stacken.kth.se>

        * nnpfs/bsd/nnpfs_vnodeops-bsd.c (nnpfs_symlink):
        vput for OpenBSD <= 3.2

2002-12-03  Love  <lha@stacken.kth.se>

        * appl/mac/mafslog/mafslog.m: 1.2.2.1->1.2.2.2: change order where
        we do k_hasafs() so we can tell use that afs client isn't started

        * appl/mac/install/Arla: handle kext's

2002-12-02  Love  <lha@stacken.kth.se>

        * appl/mac/mafslog/appl_locl.h: remove, no longer needed

        * appl/mac/mafslog/mafslog.m: inline appl_local.h so we use it
        itstead of the appl_local.h in arlalib, left over from when this
        program didn't use arlalib

        * appl/mac/mafslog/Makefile.in: no bundle for cocoa application

        * appl/mac/mafslog/Makefile.in: use libtool (for real this time)

        * appl/mac/mafslog/PkgInfo: add

        * appl/mac/mafslog/Info.plist: add

        * appl/mac/arlacmm/Makefile.in: libtool

        * appl/mac/Arla_Configuration/Makefile.in: libtool

        * HACKING: automake lib/sl/Makefile

        * lib/sl/Makefile.in: remove, use automake file

2002-12-01  Love  <lha@stacken.kth.se>

        * appl/mac/mafslog: prototype afslog

2002-11-30  Tomas Olsson  <tol@stacken.kth.se>

        * tests/run-tests.in: print date when each test is started

        * tests/build-emacs{,-j}: build --without-x, emacs-21.2

        * tests/generic-build: pass $CONFIGURE_ARGS to configure

2002-11-30  Love  <lha@stacken.kth.se>

        * arlad/messages.c: (break_callback): mirror the token managment code 
        from the kernel in here so it will be use by all users of
        break_callback()

        * arlad/fcache.c (stale): have break_callback() manaage the tokens
        when we call it, if not just clear then (since the node isn't in
        the kernel)

2002-11-29  Tomas Olsson  <tol@stacken.kth.se>

        * configure.in (vfs_mount test): include param.h before proc.h

        * doc/{debugging,oddities}.texi: speling

        * lwp/make-process.o.sh.in: speling

        * nnpfs/{aix,irix,sunos,linux,solaris}/nnpfs_node.c 
        (new_nnpfs_node): speling
        
        * arlad/inter.c (cm_open): update tokens on NNPFS_OPEN_*

2002-11-28  Love  <lha@stacken.kth.se>

        * nnpfs/include/nnpfs/nnpfs_queue.h: some useful list macros

        * arlad/fcache.c (cleaner): the 50 is really
        NNPFS_GC_NODES_MAX_HANDLE

        * ydr: generate free functions (from assar), remove tcpdump stuff

        * ydr/main.c: don't call generate_tcpdump_patches
        
        * rx/rx_pkt.h: replace rx_{Put,Get}Long with rx_Slow{Get,Put}Long
        
        * rx/rx_pkt.c (rxi_ReceiveDebugPacket): use rx_SlowPutLong
        
        * rx/rx.h: add a rock for the service to hide stuff under
        
        * rx/rx_globs.h: make rxi_Alloc return void *
        
2002-11-26  Harald Barth  <haba@stacken.kth.se>

        * conf/CellServDB: More followers from northern Sweden

2002-11-26  Mattias Amnefelt  <mattiasa@e.kth.se>

        * nnpfs/linux/nnpfs/nnpfs_node.h (1.20.2.2 -> 1.20.2.3) nodelist
        * nnpfs/linux/nnpfs/nnpfs_fs.h (1.21.2.3 -> 1.21.2.4) nodelist
        is now a hashtable
        * nnpfs/linux/nnpfs_vfsops.c (1.71.2.3 -> 1.71.2.4)
        (xfs_read_super): call xn_init_head instead of INIT_LIST_HEAD
        * nnpfs/linux/nnpfs_node.c (1.51.2.6 -> 1.51.2.7) nodelist is now
        a hashtable
        * nnpfs/linux/nnpfs_dev.c (1.77.2.6 -> 1.77.2.7) (free_node): use
        xn_deletedp
        
2002-11-18  Love  <lha@stacken.kth.se>

        * arlad/Makefile.in: build ptest, the poller test
        
        * arlad/ptest.c: poller test
        
2002-11-14  Love  <lha@stacken.kth.se>

        * conf/CellServDB (athena.mit.edu): update prill.mit.edu
        
2002-11-12  Mattias Amnefelt  <mattiasa@e.kth.se>

        * rx/rxdebug.c (MainCommand): don't change tstats.version if it's
        outside limits

2002-11-08  Tomas Olsson  <tol@stacken.kth.se>

        * nnpfs/linux/nnpfs_message.c (nnpfs_message_installdata): log
        error on data w/o token
        * .cvsignore: added compile

2002-11-05  Harald Barth  <haba@stacken.kth.se>

        * arlad/cred.c: spelling

2002-11-04  Harald Barth  <haba@stacken.kth.se>

        * configure.in: Building snapshot
        * scripts/arla-release.sh: Fix regexp and more variable quoting

2002-10-31  Love  <lha@stacken.kth.se>

        * appl/vos/vos_listaddrs.c: try to make listaddr work again
        
        * rxdef/vldb.xg: fix VL_GetAddrs, broken prototype
        
        * configure.in: more quoting for AC_PROG_EGREP, now it works for
        2.54

2002-10-30  Love  <lha@stacken.kth.se>

        * rxkad/rxk_serv.c (decode_krb5_ticket): check the enc-type of the
        encrypted part of the ticket, make sure its des
        
2002-10-29  Love  <lha@stacken.kth.se>

        * configure.in: fix comment for PROG_EGREP
        
2002-10-29  Tomas Olsson  <tol@stacken.kth.se>

        * nnpfs/winnt/bin/nnpfs-helper.c: simple logging
        * nnpfs/winnt/inc/nnpfs_fastio.h: added; fast I/O prototypes
        * nnpfs/winnt/inc/nnpfs_list.h: simple list implementation
        * nnpfs/winnt/src/sources (SOURCES): added nnpfs_fastio.c
        * nnpfs/winnt/src/nnpfs_fastio.c: added; fast I/O routines
        * nnpfs/winnt/src/nnpfs.rc: not xfs any more
        * nnpfs/winnt/src/nnpfs_{dev,dnlc,fbuf,init,message,misc,node,vops}.c: 
        write support
        * nnpfs/winnt/inc/nnpfs_{deb,dev,dnlc,ioctl,locl,node,proto}.h:
        write support

2002-10-28  Harald Barth  <haba@stacken.kth.se>

        * configure.in: This makes my autoconf complain less
        * arlad/nnpfs.c: Give better error message

2002-10-24  Harald Barth  <haba@stacken.kth.se>

        * arlad/fcache.c: Do not zero tokens if the tokens tell us that it
        is still in use from another place.

2002-10-20  Tomas Olsson  <tol@stacken.kth.se>

        * doc/nnpfs.txt: spelling
        * tests/part-files: spelling

2002-10-20  Love  <lha@stacken.kth.se>

        * arlad/fcache.c (write_attr): fix broken assert
        
        * tests/apwd.c: make readlink take size(buf) - 1, NUL terminate
        string remove unused variable
        
2002-10-19  Love  <lha@stacken.kth.se>

        * configure.in: add AC_PROG_EGREP since modern autoconf failes to
        a depend on that in AC_EGREP_CCP
        
        * lib/cmd: move to bit bucket for recycling

        * configure.in: no makefile for lib/cmd
        
        * lib/Makefile.in: don't build libcmd
        
        * appl/afsutils/Makefile.in: don't -lcmd, its not needed
        
        * conf/CellServDB: no more {cs,graphics}.cornell.edu from Mitch
        Collinsworth on arla-drinkers
        
        * util/heap.c (heap_remove): assert the we we don't try to remove
        a heap element that was inserted w/o a heap_ptr
        
2002-10-17  Love  <lha@stacken.kth.se>

        * **/*.[0-9]: format, indent, new sentence, new line

2002-10-15  Love  <lha@stacken.kth.se>

        * arlad/poller.c (poller_add_conn): use cell from conn
        
        * arlad/poller.c: remove a race when polling a server remove a
        race when adding/re-adding a entry fix a memory leak
        
2002-10-11  Mattias Amnefelt  <mattiasa@e.kth.se>

        * appl/lib/arlalib.c (arlalib_get_cred_krb) try user given cell
        before we try to guess

2002-10-08  Love  <lha@stacken.kth.se>

        * tools/release-tools/tar-rootify.c: s/<strings.h>/<string.h>/
        * tools/release-tools/cpio-rootify.c: s/<strings.h>/<string.h>/
        
2002-10-03  Love  <lha@stacken.kth.se>

        * appl/mac/make-mac-package.sh.in: look for more places where
        arla-$VERSION.pax.gz might hide
        
        * nnpfs/bsd/nnpfs_vnodeops-bsd.c (nnfs_page{in,out}): if no apple,
        #error-out

        * nnpfs/bsd/nnpfs_vnodeops-bsd.c (nnpfs_pagein): Zero out rest of
        last page if there wasn't enough data in the file
        
2002-10-02  Love  <lha@stacken.kth.se>

        * conf/CellServDB (iastate.edu): update, from Tracy Di Marco White
        <gendalia@iastate.edu>

        * lib/ko/kocell.c: fix NUL termination of fgets string
        
        * lib/cmd/cmd.c: fix NUL termination of fgets string
        
        * appl/mpp/mpp.c: fix NUL termination of fgets string
        
        * rx/rxperf.c: fix NUL termination of fgets string
        
        * arlad/fprio.c: fix NUL termination of fgets string
        
        * arlad/arla.c: fix NUL termination of fgets string
        
2002-09-30  Love  <lha@stacken.kth.se>

        * arlad/arla-cli.c: add new command, sleep

        * arlad/state.c (state_store_fcache): plug memory leak, use static
        buffer

        * arlad/state.c (state_store_volcache): plug memory leak, use static
        buffer

2002-09-30  Tomas Olsson  <tol@nada.kth.se>

        * lwp/plwp.c (Cal_Highest_runnable_priority): highest index is
        MAX_PRIORITIES - 1

2002-09-27  Love  <lha@stacken.kth.se>

        * lwp/plwp.c: more verbose debugging, set stacksize (otherwise
        uninited), make Cal_Highest_runnable_priority() work again

        * rx/rxdebug.c: replace bcopy/bzero with memcpy/memset
        
        * INSTALL: update mac os text
        
        * nnpfs/include/nnpfs/nnpfs_message.h: add more comments
        describing nnpfs_node.flags and NNPFS_TOKEN macros

        * lwp/plwp.c (lwp_init): is a pointer, use it a such
        
        * lwp/plwp.c: (Abort_LWP): this function wasn't a vararg function,
        silly, it should be, make it.
        
        * lwp/plwp.c: (LWP_CreateProcess): reorganize to make error
        handling sane, also do AbortLWP() if pthread_create/CreateThread
        failes.
        
        * lwp/plwp.c: (LWP_CreateProcess): indent

2002-09-26  Love  <lha@stacken.kth.se>

        * lwp/plwp.c (Cal_Highest_runnable_priority): fix broken code

        * util/log.c: make code common between log_open_stderr and
        log_open_file that just was in log_open_file so all fields get
        inited.

2002-09-25  Love  <lha@stacken.kth.se>

        * doc/themis.texi: more themis stuff
        
2002-09-18  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/cbdebug/cbdebug.c (apropos_cmd): check argc correctly

2002-09-17  Love  <lha@stacken.kth.se>

        * nnpfs/bsd/bin: remove depenecy on roken

        * lib/ko/koerror.c: Add some more vldb error code that I forgot
        before

        * util/log.c: set progname to "unknown-program" when we don't know
        it (wasn't passed in)

        * lib/cmd/cmd.c: use getprogname()
        
        * cf/check-sl.m4: tests for yes so we don't get -Lyes/lib & co

        * tests/*.c: use getprogname()

        * appl/lib/arlalib.c: add localauth again
        
        * appl/asrvutil: remove program

        * configure.in: remove asrvutil
        
        * appl/lib/arlalib.c: remove localauth support, I'll rewrite it soon

        * appl/lib/arlalib.c: if we use openssl, use des_random_key

        * appl/asrvutil/asrvutil.c: if we use openssl, use des_random_key
        
        * nnpfs/bsd/nnpfs_wrap-bsd.c: more more glue code to support
        netbsd current

        * nnpfs/bsd/bin/nnpfs_makedev: there might be a optional bdev-num
        
2002-09-16  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * arlad/fs_errors.h: Define __attribute__ if it's not defined and
        if not GNUC

        * util/log.h: Define __attribute__ if it's not defined and if not
        GNUC

        * util/util-tester.c: Include roken.h

        * tests/apwd.c (test_1): Don't do getcwd(NULL, 0), behaviour is
        undefined

        * lwp/make-process.o.sh.in: Try to use CPP with .S file

2002-09-14  Tomas Olsson  <tol@stacken.kth.se>

        * nnpfs/include/nnpfs/nnpfs_message.h (NNPFS_TOKEN_GOT): do things
        the traditional way, deprecate
        (NNPFS_TOKEN_GOT_{ANY,ALL}): added

        * nnpfs/include/nnpfs/nnpfs_debug.h: added XDEBREF - track
        reference counting

        * nnpfs/include/nnpfs/nnpfs_message.h (NNPFS_TOKEN_GOT): check for
        all requested tokens

2002-09-13  Johan Danielsson  <joda@pdc.kth.se>

        * nnpfs/bsd/nnpfs_wrap-bsd.c: make this compile with
        NetBSD-current

2002-09-13  Love  <lha@stacken.kth.se>

        * configure.in: add missing AC_CONFIG_SRCDIR, fix comment

        * HACKING: we need autoconf 2.53, remove exit 0, add error
        checking, call automake with -a -c to add missing files
        
2002-09-12  Love  <lha@stacken.kth.se>

        * appl/vos/vos_listaddrs.c (vos_listaddrs): make it work on cell's
        with holes in the index array

        * lwp/make-process.o.sh.in: add -DUSING_ASM to MYDEF add missing
        MYDEF spelling
        
        * configure.in: fix AC_INIT, AM_INIT_AUTOMAKE, and AC_KRB_VERSION
        
        * configure.in: catch up with new roken macros

        * cf:, lib/roken: import new roken from heimdal 0.5
        
2002-09-12  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/afsutils/Makefile.in: Don't install klog

        * arlad/fcache.c (cleaner): if state is CL_COLLECT, bail out
        unconditionally

        * nnpfs/linux/nnpfs_message.c (nnpfs_force_invalid_xnode): do
        invalidate_inode_pages

2002-09-12  Love  <lha@stacken.kth.se>

        * configure.in: check for <limits.h>
        
        * nnpfs/linux/Makefile.in: catch up with `pacify automake' changes
        
        * configure.in: prepare for new libroken, pacify automake
        
        * rxkad/Makefile.in: don't misuse VPATH, use srcdir
        
        * xfs: xfs is dead, long live nnpfs

2002-09-11  Harald Barth  <haba@pdc.kth.se>

        * nnpfs/linux/nnpfs_dev.c: Place down(), the enter of the mutex,
        earlier and move asignments which lingered in the variable
        declaration area inside the area protected by the mutex (compare
        with the corresponding solaris code). Use the newly assigned value
        of sleepq in the first part of the for(;;) construct to init the
        loop variable.

2002-09-09  Mattias Amnefelt  <mattiasa@e.kth.se>

        * configure.in: Try to find linux includefiles in a way consistent
        with modern linux kernels.

2002-09-08  Love  <lha@stacken.kth.se>

        * nnpfs/bsd/nnpfs_message.c (nnpfs_message_invalidnode): in the
        __APPLE__ case, vget the node before calling ubc functions
        
        * nnpfs/bsd/nnpfs_vnodeops-common.c (nnpfs_handle_stale): set ubc
        size when invalidating the node

        * doc/nnpfs.txt: ispelling and technical updates
        
        * **/*: 's/nnpfsserver/xfsserver/g'

        * nnpfs/bsd/nnpfs_message.c: indent
        
2002-09-08 Harald Barth <haba@stacken.kth.se>

        * doc/nnpfs.txt: Lots of changes, please review.
        
2002-09-08  Mattias Amnefelt  <mattiasa@e.kth.se>

        * nnpfs/linux/nnpfs_inodeops.c (nnpfs_d_revalidate): added debug
        message

2002-09-08  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * configure.in: Added --with-milkopart

        * lib/ko/afs_uuid.c: Include sys/file.h if it exists

        * configure.in: Added sys/mkdev.h to AC_CHECK_HEADERS

2002-09-07  Love  <lha@stacken.kth.se>

        * configure.in: require autoconf 2.53
        
2002-09-07  Mattias Amnefelt  <mattiasa@e.kth.se>

        * lib/ko/kocell.c (cell_init): read cells from $CELLSERVDBFILE if
        set. also protect suid-binaries

        * appl/udebug/udebug.c: print alternate addresses
        * rxdef/ubik.xg: added alternate addresses in ubik_sdebug

        * appl/bos/bos.8 (getrestart) updated

        * appl/fs/fs_setcrypt.c: added help

2002-09-07  Love  <lha@stacken.kth.se>

        * arlad/messages.c: Always set offset in the installdata
        message. Noticed by <tol@it.su.se>.

        * arlad/fcache.c: adding missing ) and spelling, From Fredrik Thulin
        <ft@it.su.se>

        * nnpfs/linux: handle the changed semantics of ->d_revalidate

        * nnpfs/irix/nnpfs_vfsops.c: s/xFs/nnpfs/, close that I missed
        this one :)

        * **/* (with exception of ChangeLog,NEWS and timeline.texi):
        s/xfs/nnpfs/g;s/XFS/NNPFS/g
        
        * nnpfs: repo copy of xfs
        
        * xfs: rename to nnpfs

        * arlad/stats.c (collectstats_hostpart): remove gracious extra
        varible

        * arlad/messages.h: remove FETCH_BLOCKSIZE
        
        * arlad/stats.c: define MIN_FETCH_BLOCKSIZE here and
        s/FETCH_BLOCKSIZE/MIN_FETCH_BLOCKSIZE/

        * arlad/stats.c (collectstats_hostpart,collectstats_init): use
        growing hashtab

        * arlad/stats.c (hostpart_addhash): use hashtabaddreplace() to
        avoid memory leak

        * arlad/message.[ch],stats.[ch]: enable prefetch again

        * arlad/messages.c (xfs_message_putattr): don't update length at
        the same time as we update attributes, since the fileserver can't
        handle that.

2002-09-06  Love  <lha@stacken.kth.se>

        * configure.in: fix netbsd-lkm with relative $srcdir
        
2002-09-05  Love  <lha@stacken.kth.se>

        * arlad/stats.c (collectstats_stop): don't count statistics if
        partition is -1

2002-08-29  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * xfs/bsd/Makefile.in: Changed Mac OS X to use kext:s

        * xfs/bsd/Info.plist.in: Added

        * configure.in: Added COM_APPLE_KERNEL_BSD and xfs/bsd/Info.plist
        generation

2002-08-26  Love  <lha@stacken.kth.se>

        * tools/release-tools/Makefile.in: don't install/uninstall
        
2002-08-25  Love  <lha@stacken.kth.se>

        * conf/CellServDB: su.se does afsdb, so should you
        
2002-08-24  Love  <lha@stacken.kth.se>

        * lib/bufdir/fbuf.c: Don't map the whole file in mmap_copy.*
        functions, do instead chunks of mmap_max_size (10M). This way
        operating system that have problems to maps things larger then
        600M will still work.
        
2002-08-21  Love  <lha@stacken.kth.se>

        * xfs/linux/xfs_syscalls.c (sys_afs_int): simplify len logic in
        the copyout case
        
        * xfs/linux/xfs_inodeops.c: (xfs_d_delete): fix comment

2002-08-17  Love  <lha@stacken.kth.se>

        * rxkad/Makefile.in: build fc_test
        
2002-08-16  Love  <lha@stacken.kth.se>

        * arlad/messages.c (token_for_cell): use XFS_MSG_MAX_DATASIZE for
        buffer size

        * arlad/messages.c (viocsettok): more error checking
        (token_for_cell): more error checking

        * appl/lib/tokens.c (arlalib_token_iter): more error checking
        
2002-08-15  Love  <lha@stacken.kth.se>

        * Release 0.35.9, see the branch for relevant patches
        
2002-08-14  Love  <lha@stacken.kth.se>

        * NEWS: sync with 0.35 branch
        
2002-08-13  Love  <lha@stacken.kth.se>

        * arlad/arla.c (arla_init): if the user choose "syslog" as
        log_file, change it to "syslog:no-delay" to be helpful since that
        is probably what the user wants
        
        * NEWS: sync in changes from 0.35.9
        
        * xfs/aix/xfs_syscalls.c (xfs_pioctl_call): use
        XFS_MSG_MAX_DATASIZE

        * xfs/solaris/xfs_syscalls.c (remote_pioctl): use
        XFS_MSG_MAX_DATASIZE

        * xfs/bsd/xfs_syscalls-common.c (remote_pioctl): use
        XFS_MSG_MAX_DATASIZE

        * xfs/linux/xfs_syscalls.c (sys_afs_int): use XFS_MSG_MAX_DATASIZE
        
        * xfs/include/xfs/xfs_message.h (XFS_MSG_MAX_DATASIZE): new
        constant
        (xfs_message_pioctl, xfs_message_wakeup_data): use XFS_MSG_MAX_DATASIZE

        * include/kafs.h: add common pioctl's
        
2002-08-10  Love  <lha@stacken.kth.se>

        * conf/CellServDB: add lns.mit.edu
        
2002-08-09  Love  <lha@stacken.kth.se>

        * cf/*: do offical workaround for LTLIBOBJS
        
        * configure.in: add AC_PROG_EGREP for those that have it do
        LTLIBOBJS wordaround as documented
        
2002-08-05  Love  <lha@stacken.kth.se>

        * rxdef/vldb.xg: set limit to bulkaddrs<>
        
        * ydr/output.c (gen_check_overflow): check for overflows
        (encode_string):check for length -1 that would break the code otherwise
        (encode_varray): use gen_check_overflow

        * lib/Makefile.in: no libacl for now
        
        * rxdef/volumeserver.xg: sync with openafs s/replicas/manyDests/,
        put a limit on it too
        
        * rxdef/volumeserver.xg: AFSVolForwardMultiple/multi_results is
        OUT

2002-08-01  Love  <lha@stacken.kth.se>

        * arlad/messages.c (xfs_message_create): handle tokens for
        installnode
        (xfs_message_mkdir): ditto
        (xfs_message_getdata): don't allow to open directories with DATA_W
        Noticed by <tol@stacken.kth.se>

2002-07-27  Love  <lha@stacken.kth.se>

        * arlad/fcache.c (purge_cred): use FCACHE2XFSNODE_NO_LENGTH
        
        * arlad/messages.c: move the fake_stat code into
        afsstatus2xfs_attr(), obey the FCACHE2XFSNODE_LENGTH flag in
        afsstatus2xfs_attr(). this the had the effect that token changes
        will update rights when running with fake_stat

        * arlad/messages.h: change the FCACHE2XFSNODE flags to be more
        sane, the one that we really what make a make is possible to avoid
        is update the length of the file, so make that a flag instead
        
        * appl/mac/make-mac-package.sh.in: added more descriptive helptext,
        added Id tag
        
        * tests/intr-read[123]: intr-read is not a fast test

        * conf/CellServDB (in2p3.fr): update
        * conf/CellServDB (rl.ac.uk): update
        
2002-07-26  Love  <lha@stacken.kth.se>

        * tools/release-tools/common.c: use strtoq if there is no strtoll
        use strol if there is no strtoq
        
        * configure.in: check for strtoq

        * arlad/fcache.c (do_read_attr): print out warning with error
        before setting error to ENETDOWN

        * appl/mac/make-mac-package.sh.in: post-process with release-tools
        
2002-07-25  Love  <lha@stacken.kth.se>

        * lib/ko/kocell.c (cell_new_dynamic): remove write-to-CellServDB
        code

        * lib/ko/Makefile.in: added commented out ares support
        
        * appl/perf/afsfsperf.c: add -smallrun, spelling, help-tests From:
        Harald Barth <haba@pdc.kth.se>
        
2002-07-24  Love  <lha@stacken.kth.se>

        * lib/ko/restest.c: memset dbservers, print address, fix usage of
        _ko_resolve_host

        * lib/ko/ares.c: implementation of a libares resolver backend, need
        some patches to work
        
        * lib/ko/kocell.c (dns_lookup_cell): don't add time to the tll for
        hosts that have invalid timeout/address since that will render the
        address (0.0.0.0) valid
        
        * arlad/arladeb.h: unbreak (c)
        
2002-07-24  Tomas Olsson  <tol@stacken.kth.se>

        * xfs/winnt/bin/xfs-helper.c (device_thread, loop): cleaner debug
        output, don't fail when ioctl says ENOMEM DEBUG off by default

2002-07-24  Love  <lha@stacken.kth.se>

        * lib/ko/resolve.c: simplify, fixed some minor bugs, more strcasecmp
        
        * lib/ko/reswinnt.c: first shot of a windows <windns.h> resolver
        for libko

        * lib/ko/Makefile.in: restest
        
        * lib/ko/restest.c: restest

        * lib/ko/kocell.c (dns_lookup_cell): make lowest ttl 5 minutes,
        move comment

        * lib/ko/resnull.c: Dummy implementation of resolver api
        
        * lib/ko/Makefile.in: add RESOLVE_[CO]
        
        * lib/ko/resolve.c: Implementation of resolver api using roken
        resolve.h moved from kocell.c
        
        * lib/ko/kocell.c: use the independant resolver api to resolve
        cell/hosts

        * lib/ko/ko_locl.h: create a resolver independant resolver api for
        libko

2002-07-23  Tomas Olsson  <tol@stacken.kth.se>

        * arlad/xfs.h (xfs_message_receive): remove duplicate

        * xfs/winnt/src/* handle symlinks, reset cache, ...
        
        * xfs/winnt/inc/xfs_proto.h (xfs_alloc{,_link}): new arg - tag

        * xfs/winnt/inc/xfs_node.h (xfs_node_{gc_all,invalid}): added

        * xfs/winnt/inc/xfs_locl.h (xfs_node): reference counts are now long
        (xfs_channel): add wake_event
        define XFS_MAX_SYMLINKS

2002-07-23  Love  <lha@stacken.kth.se>

        * arlad/fcache.c (find_volume): only tell us to go get the rw
        volume if there exists one

        * arlad/Makefile.in: bring uptodate
        
        * arlad/volcache.c: push the disconnected test to later in the
        code so dynroot volumes lookups will work
        (volume_uptodatep): the volume is only valid in the disconnected
        case when we have valid data
        
        * arlad/dynroot.c: add aliases to the generated dynroot
        
        * lib/ko/kocell.c: add support for aliases in dynroot
        
        * lib/ko/ko.h (cell_alias_foreach): new function
        (cell_addalias): new function

        * arlad/arla.c (params): sort and add fetch_block
        
        * arlad/inter.c: indent
        
        * arlad/dynroot.h: update (c)
        
        * arlad/{darla.c,darla.h,discon.h,discon_log.c,discon_log.h,
        reconnect.c,reconnect.h}: mark obsolete

        * arlad/stats.c (stats_set_prefetch): no return from void
        
        * arlad/stats.c (stats_set_prefetch): new function, set prefetch
        variable
        (stats_prefetch): return fetch_block_size instead of fixed size

        * arlad/stats.h (stats_set_prefetch): new function, set prefetch
        variable

        * arlad/arla_local.h: stop include unused files export
        fetch_block_size
        
        * arlad/*.[58]: update, improve, add more junk

2002-07-22  Love  <lha@stacken.kth.se>

        * configure.in: check for lchflags too
        
2002-07-21  Love  <lha@stacken.kth.se>

        * arlad/inter.c (cm_check_consistency): fcache_calculate_usage now
        correctly returns a int64_t

        * arlad/fcache.h (fcache_calculate_usage): return a int64_t
        * arlad/fcache.c (fcache_calculate_usage): return a int64_t
        
        * arlad/fcache.c (write_data): fix byte accounting, we leaked
        bytes here before

2002-07-20  Love  <lha@stacken.kth.se>

        * arlad/dynroot.c (dynroot_get_node): use propper fcache function
        to update accunting length
        
        * arlad/fcache.c (read_data): decide how much read-ahead using the
        function stats_prefetch()
        (sum_node): include dynroot nodes again

        * arlad/stats.h: warpping and prototype for stats_prefetch()
        
        * arlad/stats.c (stats_prefetch): return how much readahead we
        should do for this fileserver, now wired to FETCH_BLOCKSIZE
        
        * arlad/arla_local.h: add "stats.h"

        * arlad/fcache.h: move out collectstats_ functions to stats.h
        
        * arlad/{Makefile.in.stats.[ch].fcache.c}: break out statistics
        counting to own module

        * arlad/fcache.c (cleaner): removed unused debugging code
        
        * INSTALL: --with-krbafs, point toward README
        
        * xfs/bsd/Makefile.in: add a install-native that tries to do the
        right thing make lkmdir mean lkmdir (almost at least)
        (same for uninstall)

        * xfs/bsd/bin/Makefile.in: make lkmdir mean lkmdir (almost at
        least)

2002-07-17  Mattias Amnefelt  <mattiasa@e.kth.se>

        * arlad/disco.c (disco_print_entry) terminate default with break

2002-07-16  Love  <lha@stacken.kth.se>

        * appl/vos/vos_common.c: fix memset arguments, from openbsd
        
        * include/Makefile.in (LOCL_HEADERS): split out roken headers
        
        * configure.in: provide am conditional ARLA_LOCAL_ROKEN that can
        be used in makefiles

2002-07-15  Love  <lha@stacken.kth.se>

        * appl/mpp/mpp.c: fix expansion of macros
        
        * tests/*.c: s/set_progname/setprogname/

        * xfs/bsd/xfs/xfs_node.h (xfs_node): lockf should me a pointer to
        a struct lockf

        * lib/cmd/testc.c: s/set_progname/setprogname/
        
2002-07-14  Love  <lha@stacken.kth.se>

        * configure.in: KERNEL_FUNCS+= lf_advlock

        * xfs/bsd/xfs_vnodeops-bsd.c, xfs/bsd/xfs/xfs_node.h: do local
        locking if possible

2002-07-11  Love  <lha@stacken.kth.se>

        * scripts/arla-release.sh: manage current arla version extracting
        too (AM_INIT_AUTOMAKE vs VERSION)

        * appl/afsutils/klog.c: rewrite the code even more, this sucks
        
        * appl/afsutils/klog.c: fix for finding of ticket based on
        proposal from Christopher Wing <wingc@engin.umich.edu> on
        arla-drinkers, changed by me.

2002-07-11  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/mac/make-mac-package.sh.in: Make script take srcdir and
        destdir

2002-07-10  Love  <lha@stacken.kth.se>

        * arlad/reconnect.c: s/rx_Error/rx_GetCallError/
        * arlad/fcache.c: s/rx_Error/rx_GetCallError/
        * lib/bufdir/fbuf.c: s/rx_Error/rx_GetCallError/
        
        * arlad/xfs.c: move rpc messages counters to a struct to make
        easier to add new counters

2002-07-09  Love  <lha@stacken.kth.se>

        * appl/vos/vos_{examine,vldbexamine}.c: fix printing when
        volserdown, broken printing.

2002-07-08  Tomas Olsson  <tol@stacken.kth.se>

        * xfs/winnt/bin/xfs-helper.c (device_thread): proper bailout on error
        (loop): repeat recv until we have the whole packet

        * xfs/winnt/inc/xfs_{dnlc,errno,locl,node,proto}.h:
        read some more

        * xfs/winnt/src/xfs_{deb,dev,dnlc,fbuf,init,message,misc,node,vops}.c:
        read some more

2002-07-05  Love  <lha@stacken.kth.se>

        * NEWS: mention afs3-callback
        
        * arlad/arlad.c (client_port): default to 0, and thus to
        afscallbackport.

        * arlad/arla.c (rxinit): if no port is specified use
        afscallbackport

2002-07-04  Love  <lha@stacken.kth.se>

        * arlad/fcache.c (cleaner): Change where we update the cleaner run
        number of the node to just after we check for it. Each change of
        state in the cleaner will change the cleaner run number so that
        each state look at every node. Fixes bug noticed by
        <tol@stacken.kth.se>.

        * arlad/fcache.c (cleaner): always consider a node cleanable if
        its used in the first step, since we never clean used used nodes
        later in the code.
        (invalidator): use ADEBINVALIDATOR
        (read_data): fix needbytes calculation

        * arlad/arladebu.c (arla_deb_units): add ADEBINVALIDATOR
        * arlad/arladeb.h: masks: add ADEBINVALIDATOR
        
2002-07-03  Love  <lha@stacken.kth.se>

        * appl/perf: add from milko appl tree, clean up

        * configure.in (AC_OUTPUT): move perf form milko appl to arla appl
        
        * doc/xfs.txt (version): clean up
        (installdata): make more sane

        * conf/freebsd/freebsd4-rc.arla: freebsd 4 start script
        (/etc/rc.arla)

2002-07-02  Love  <lha@stacken.kth.se>

        * appl/mpp/mpp.c (expand_macro): fix a broken case where there was
        multiple macros on the same row
        
2002-07-01  Love  <lha@stacken.kth.se>

        * arlad/arla.c (read_conffile): make negative boolan values work
        
        * arlad/messages.c: (try_next_fs): protect RXKADUNKNOWNKEY with
        #ifdef KERBEROS, from tol

        * arlad/cmcb.c: (RXAFSCB_GetCellServDB): fix prototype, from tol

        * arlad/arla.c: protect rxkad_level_string with #ifdef KERBEROS
        
2002-06-30  Love  <lha@stacken.kth.se>

        * arlad/conn.c (conn_dead): tell in what cell the dead host is in
        
        * tests/run-tests.in: RUN_TESTS += intr-read2 intr-read3
        
        * tests/intr-read{.c,1,2,3}: sigchild test

        * xfs/bsd/xfs_dev-common.c (xfs_block_sigset): ignore dying children
        (xfs_message_rpc): correct comment

2002-06-29  Love  <lha@stacken.kth.se>

        * arlad/messages.c (xfs_message_open): add tokens to e->tokens,
        mirrors cm_open

        * xfs/bsd/xfs_vnodeops-netbsd.c (xfs_netbsd_strategy): don't do
        sync io

        * arlad/inter.c (cm_rename): when renaming a file we also need to
        check if the file is in the cache, since in that case we need to
        mark it as a silly rename in case its opened by the kernel, and
        thus need to write it back to the server (where it no longer
        exists!)

2002-06-28  Johan Danielsson  <joda@pdc.kth.se>
        
        * appl/lib/tokens.c: don't bail out before we get EDOM (signaling
        the end of the tokens), the kernel can also return ENOTCONN,
        meaning that the index does not exist anymore (for example if the
        token has expired)
        
2002-06-28  Love  <lha@stacken.kth.se>

        * Release 0.35.8, see the branch for relevant patches

2002-06-27  Love  <lha@stacken.kth.se>

        * tools/Makefile.in (SUBDIRS): add release-tools
        
2002-06-25  Love  <lha@stacken.kth.se>

        * lwp/process.sparc.S: rename AS_UNDERSTANDS_REGISTER to
        PROG_AS_UNDERSTANDS_REGISTER
        
        * configure.in: rename AS_UNDERSTANDS_REGISTER to
        PROG_AS_UNDERSTANDS_REGISTER

2002-06-24  Love  <lha@stacken.kth.se>

        * README: update where krbafs lives
        
2002-06-22  Love  <lha@stacken.kth.se>

        * xfs/linux/xfs_inodeops.c (xfs_vma_close): don't iput
        
        * lib/ko/afs_uuid.c (get_node_addr): just use dl->sdl_type
        directly instead of fooling around with struct if_data
        
        * conf/arla.conf: add rxkad-level example
        
        * lib/ko/kocell.c: remove direct dep on libroken (emalloc/estrdup)
        
2002-06-14  Love  <lha@stacken.kth.se>

        * ydr/output.c: use rx_SetCallError and rx_GetCallError

        * rx/rx.h: add rx_SetCallError and rx_GetCallError

2002-06-13  Love  <lha@stacken.kth.se>

        * arlad/arla.conf.5: sort
        
        * arlad/arla.c: add rxkad-level, sort

2002-06-12  Love  <lha@stacken.kth.se>

        * arlad/messages.c: (fcacheentry2xfsnode): add hook for fake_stat

        * arlad/arlad.c: (args): add --fake-stat and sort
        
        * arlad/arla.c,arla_local.h: add fake_stat

        * arlad/cb.xg: add GetCacheConfig

        * arlad/cmcb.c: add RXAFSCB_GetCacheConfig

        * arlad/poller.c: fix comments

2002-06-12  Mattias Amnefelt  <mattiasa@e.kth.se>

        * conf/CellServDB: update msc.cornell.edu

2002-06-06  Mattias Amnefelt  <mattiasa@e.kth.se>

        * appl/udebug/udebug.c (ProbeHost) Don't try to talk to
        0.0.0.0. It won't answer.

2002-06-06  Love  <lha@stacken.kth.se>

        * appl/lib/fs_lib.c (getaddrinfo): don't return a negativ error
        
        * lwp/lwp_asm.c: s/int32_t/unsigned long/
        
2002-06-05  Love  <lha@stacken.kth.se>

        * lwp/lwp_asm.c: don't used strlcpy, we don't want a dependency on
        libroken
        
2002-06-05  Mattias Amnefelt  <mattiasa@e.kth.se>

        * appl/cbdebug/{localcell,whoareyou}.c: noauth is a flag

2002-06-05  Love  <lha@stacken.kth.se>

        * arlad/cmcb.c (RXAFSCB_GetLocalCell): implement
        
        * appl/cbdebug/cbdebug.1: document localauth and update usage
        
        * appl/cbdebug: clean up, merge common code, and add localcell

        * rxdef/cb.xg: add GetCellServDB and GetLocalCell
        
2002-06-04  Love  <lha@stacken.kth.se>

        * rx/rxperf.c: if we have getrusage and there is a -G, print
        rusage stats

        * configure.in: add check for getrusage
        
        * rx/rxperf.c: comment out rx_window, it doesn't belong here
        
        * lwp/Makefile.in: lwp redzone
        
        * configure.in: Add hook to turn off lwp redzone
        
        * lwp/lwp_asm.c: support those OS need to have have a fd to
        /dev/zero when mapping anonymous pages (ie those that doesn't have
        MAP_ANON)

        * TODO: the arlad bugs are fixed and verified to no occur, remove,
        now only xfs item left
        
        * arlad/fcache.c (throw_entry): make sure there isn't waiters for
        this lock, since we don't handle that case right now
        
2002-06-03  Love  <lha@stacken.kth.se>

        * configure.in: add aafs and release-tools
        
        * tools/release-tools/Makefile.in: new file
        
        * tools/release-tools: make compile cleanly
        
        * tools/release-tools: add cpio-rootifier and break out common stuff

        * lwp/plwp.c: remove def of PRE_Block
        
2002-06-03  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * lib/ko/afs_uuid.c (get_node_addr): check if ifdata is NULL

2002-06-03  Love  <lha@stacken.kth.se>

        * tools/release-tools/tar-rootify.c: hack to root-ify tar archives

        * aafs/aafs-perl: perl interface to aafs, need some more work

        * appl/aafs: fetch info from vldb/volserver server library, more
        coming soon

        * lwp/process.sparc.S: make it work for netbsd/sparc64 (really any
        sparcv9 running !solaris)

        * configure.in (netbsd): make for-loop into a while test $#
        .. loop so we can parse two component items (-isystem)
        
        * configure.in (netbsd): make mkinstalldirs quiet
        
        * INSTALL: add notice about source-code for *BSD and remove text
        about netbsd 1.4.2 brokenness
        
        * configure.in: remove the need to copy the foo.c in the netbsd
        buildlink thingy s/with-make/with-bsd-make/
        
        * README: clean up versions supported
        
        * INSTALL: add --with-bsd-make
        
        * configure.in: --with-make, s/dir/make/
        
        * include/netbsd-lkm: buildlink-env for netbsd lkm's
        
        * configure.in: provide infrastructure for creating a build-link env in
        include/netbsd-lkm for netbsd kernel modules
        
2002-06-02  Love  <lha@stacken.kth.se>

        * netbsd 1.6 should also do the nm trick

        * arlad/arla.conf.5: document the broken parse_units
        
        * doc/pioctl.texi: document AIOC_GETCACHEPARAMS
        
        * arlad/messages.c: implement support for AIOC_GETCACHEPARAMS
        
        * appl: try to support 64-bit cache parameters

        * include/kafs.h: add AIOC_GETCACHEPARAMS and its opcodes
        
        * arlad/arla.c (read_conffile): parse_units returns a int, and
        wraps over, assume that negative returns are such overwrapped
        values.

        * arlad/arla.c (read_conffile): use strtoll if we have it
        
        * arlad: first part of support of large cache sizes (>4G)

        * configure.in: check for <sys/wait.h> we need it if we link
        against external roken

        * lwp/rw.c: try harder to not store interger in pointer
        
        * lwp/lwp_asm.c: LWP_REDZONE depends on HAVE_MMAP, make it so
        
        * lwp: revert the PRE_Block change, it not correct either way, and
        it breaks least process.S things by having it a char

        * xfs/bsd/xfs_vnodeops-netbsd.c (xfs_netbsd_read): uio_resid might
        be long, cast appropriately

        * arlad: const poisoning

        * arlad/fcache.c: update users of volcache_getby{id,name} to pass
        in NULL as type when it isn't need later in the code
        
        * arlad/volcache.c (create_new_entries): use malloc
        (add_clone): remove the suffix type, its no longer used
        (*): use dynroot_fetch_root_vldbN, update uses of add_clone
        (volcache_getbyname): make it voluntary to send in type
        (get_info_loop): pass in the entry to update, not the existing one

        * arlad/dynroot.[ch] (dynroot_fetch_root_vldbN): rename from
        dynroot_fetch_vldbN()

        * arlad/arlad.c: indent
        
2002-06-01  Love  <lha@stacken.kth.se>

        * lwp: remove register, ident some
        make PRE_block an int since some arch .S files access it as such
        add some comment to LWP_REDZONE code and rename some
        functions/variables, also pass in more info to free (since we know it)
        constify some functions
        add test-cases for the redzone code
        
        * conf/CellServDB: remove {mr-afs,blubb}.pdc.kth.se
        
2002-05-31  Love  <lha@stacken.kth.se>

        * rxdef: stop include stuff we dont need

        * arlad/messages.c (CONNMODE_FETCH): open disco log
        
        * appl/cbdebug/whoareyou.c: include more stuff
        
        * lib/ko/afs_uuid.c: include more stuff
        
        * rx: s/MUTEX_.*\>/RX_MUTEX_\1/ to make sure we dont pollute namespace

2002-05-31  Mattias Amnefelt  <mattiasa@e.kth.se>

        * appl/vos/vos_listaddrs.c: added -uuid and -host flags
        * appl/lib/arlalib.{c,h}: added arlalib_name_to_host

2002-05-31  Love  <lha@stacken.kth.se>

        * doc/timeline.texi: mention that the darwin codes originates from lxs
        
2002-05-30  Love  <lha@stacken.kth.se>

        * rxdef/vldb.xg: add some comments
        
2002-05-30  Mattias Amnefelt  <mattiasa@e.kth.se>

        * rxdef/vldb.xg: added new returncodes from openafs
        * lib/ko/afs_uuid.{c,h} (afsUUID_{hash,from_string,is_nil}) const args
        * appl/vos/vos_listaddrs.c: added, implements listaddrs
        * appl/vos/vos.{8,c} appl/vos/vos_local.h: added listaddrs
        
2002-05-28  Love  <lha@stacken.kth.se>

        * arlad/volcache.c (update_entry): common code between
        volcache_getbyname, volcache_getbyid and volume_make_uptodate
        
        (add_clone): add entry even if suffix_type doesn't match type
        (remove_clones_from_hashtab): new func
        (remove_clone): new func
        (add_clones_to_hashtab): comment out code that in spirit is right
        (get_info_common): make it take an nvldb entry that will be used when
        checking the entry out
        (volume_make_uptodate): make it do the correct thing

        * lib/ko/kocell.c (updatehosts): where there is no update on the
        dbservers array, don't copy it to a new malloced area and free the
        old.
        (fetch_host): hosts returned by gethostbyname() are valid for 600s.

        * TODO: some things fixed, some things doesn't apply
        
        * arlad/fcache.c
        (fcache_reobtain_callbacks,fcache_giveup_all_callbacks): avoid
        talking to dead hosts

        * arlad/arla.conf.5: fix dynroot example
        
        * conf/arla.conf: fix dynroot example
        
        * arlad/arla.c (read_conffile): support boolan values
        (conf_params): make dynroot a boolan value

2002-05-27  Love  <lha@stacken.kth.se>

        * rxdef/vldb.xg: clean up and add opcodes
        
        * doc: rename disco-prog.texi to prog-disco.texi

2002-05-27  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/fs/fs.c: Spell-checked command descriptions

2002-05-26  Love  <lha@stacken.kth.se>

        * doc/disco-prog.texi: clean up
        
        * doc/prog.texi: add afsUUID
        
        * doc/arla.texi: add afsUUID
        
        * include/kafs.h (CONNMODE_CONN_WITHCALLBACKS): add
        
        * arlad/d-trans.c,disco.[ch],play_disco.c: disco stuff

        * arlad/messages.c (*): use disco to store log entries, remove
        log_dis stuff, introduce CONNMODE_CONN_WITHCALLBACKS, so that we
        reconnect w/o getting callback again
        (VenusFid_cmp): move to libko

        * arlad/fcache.c (remove_file): add support for disco
        
        * arlad/fcache.h (FCacheEntry): add disco_id
        
        * lib/ko/Makefile.in: add misc and add EXEEXT hack for gensysname
        
        * lib/ko/misc.c: misc stuff, currently VenusFid_cmp that is shared
        between aafs and arlad
        
        * doc: clean up, add docu on new stuff, split of programming stuff in to own section
        
        * doc/real-world.texi: add note about kimpersonate
        
2002-05-25  Love  <lha@stacken.kth.se>

        * arlad/poller.c (poller_add): make time timeout stuff work again
        
        * arlad/poller.c (poller_add): check if the cell has a special
        timeout value, in that case we might want to use it (think NAT)
        
        * lib/ko/kocell.c: add volume-type printing stuff here
        
        * lib/ko/ko.h: add volume-type printing stuff here add poller
        timeout value to cell
        
        * appl/vos: move volume-type printing stuff to libko

        * configure.in: de-support hack for netbsd 1.4.1
        
        * xfs/bsd/xfs_vnodeops-common.c (xfs_fetch_rights): remove
        (xfs_attr_valid): make it check the pag too
        (xfs_access_common): stop calling xfs_fetch_rights, xfs_attr_valid
        does the same things
        
        * xfs/bsd/xfs_vnodeops-netbsd.c (xfs_netbsd_strategy): use already
        figured out curproc and only update the vattr if the write/read
        succeeded

2002-05-24  Love  <lha@stacken.kth.se>

        * appl/vos/vos_listvldb.c (vos_listvldb_iter): check error value,
        init variables, and even better, free used memory. Bug found by
        <jimmy@e.kth.se>

        * arlad/arla.c (read_conffile): compare endptr (from strtol) with
        '\0' since that is that since of that there wasn't any bad things
        in the string.  From: Fredrik Thulin <ft@it.su.se>

        * rxdef/vldb.xg (VLSF_NEWREPSITE): dunno what this is ???
        
        * arlad/reconnect.c (copy_cached_file): open with mode, target
        file should be opened with O_TRUNC, from Andreas Östling
        <andreaso@it.su.se>
        
2002-05-23  Love  <lha@stacken.kth.se>

        * util/hash.c (hashtabnewf): size of zero is valid if the hash is
        a growing hashtable, remember to store the flags in the newly
        created hashtable
        (_might_resize): remeber to copy the size too

        * arlad/dynroot.c (DYNROOT_ROOTVOLUME_STR): string version of
        DYNROOT_ROOTVOLUME
        (dynroot_isvolumep): use DYNROOT_ROOTVOLUME_STR

        * rx/rxperf.c: clean up comments and usage
        
2002-05-22  Love  <lha@stacken.kth.se>
        
        * (free_fs_server_context): only process marks if there was a
        volume entry to process marks in, matters when in disconnected
        mode and we failed to talk to the fileserver/volumeserver.
        
2002-05-21  Love  <lha@stacken.kth.se>

        * doc/ack.texi: update
        
        * arlad/fcache.c (fcachecmp): use VenusFid_cmp when comparing fids
        
        * arlad/cmcb.h (cmcb_reinit): new interface
        
        * arlad/cmcb.c (cmcb_reinit): new interface that get you a new
        afsUUID
        (cmcb_init): use cmcb_reinit
        (RXAFSCB_CallBack): use poller_host2cell

2002-05-18  Love  <lha@stacken.kth.se>

        * arlad/poller.c (poller): don't talk to !conn_alive hosts, assume
        they are dead

        * arlad/arla.conf.5: added dynroot
        * conf/arla.conf: add dynroot
        
2002-05-17  Love  <lha@stacken.kth.se>

        * lib/ko/ko_locl.h: added <limits.h>, needed for standards
        complient platforms :)

        * arlad/messages.c (viocvenuslog): print cell status
        
        * lib/ko/kocell.c: rename cell_info to cell_status, like other
        modules
        * lib/ko/ko.h: rename cell_info to cell_status, like other modules
        
        * arlad/messages.c (xfs_message_getdata): use message_get_data
        when perfetching
        
2002-05-17  Harald Barth  <haba@pdc.kth.se>

        * arlad/messages.c (xfs_message_getdata): Fetch offset as 
        requested from the read instead of taking a random value 
        and bracketing it with prefetchsize and filesize.

2002-05-17  Love  <lha@stacken.kth.se>

        * appl/lib/arlalib.c (arlalib_get_viceid_servers): make sure res
        is inited

        * lib/ko/kocell.c (dns_lookup_cell): init lowest_ttl to INT_MAX
        and lets catch it in the end of the function.
        
        * lib/ko/kocell.c (cell_info): print all cell info
        
        * lib/ko/ko.h: add cell_info()
        
        * doc/pioctl.texi: add types of data for
        STATISTICS_OPCODE_GETENTRY

        * appl/fs/fs_statistics.c (stat_types): order types
        
        * appl/fs/fs.1: getstatistics docu
        
        * appl/fs/fs_statistics.c: enable stats on fetch status, and while
        i'm here, have a table of name and value to have one place where
        they exist

        * arlad/fcache.c (write_attr): store stats on storestatus
        
        * include/kafs.h: add STATISTICS_REQTYPE_STORESTATUS
        
2002-05-17  Hans Insulander  <hin@stacken.kth.se>

        * **/*: diffrent -> different

2002-05-16  Jimmy Engelbrecht  <jimmy@stacken.kth.se>

        * appl/fs/fs_statistics.c: statisticscode does now care about writes
        * arlad/fcache.c: statisticscode does now care about writes
        * include/kafs.h: statisticscode does now care about writes
        
2002-05-16  Love  <lha@stacken.kth.se>

        * arlad/messages.c (message_get_data): pass in wanted_length so we
        can use the function for files that we don't want to fetch the
        whole file.
        
2002-05-16  Hans Insulander  <hin@stacken.kth.se>

        * **/*.[0-9]: Lots of manpage nitpicking. More to come the next
        time i'm bored...

2002-05-16  Love  <lha@stacken.kth.se>

        * xfs/bsd/xfs_node-bsd.c (free_all_xfs_nodes): FreeBSD and OpenBSD
        uses diffrent semantics, FreeBSD have does vrele and OpenBSD does
        vgone.

        * xfs/bsd/xfs_wrap-bsd.c: print less cryptic message when loading
        kernel module, also print out arla and xfs protocol version. From
        idea of Lex Wennmacher <wennmach@netbsd.org>
        
        * xfs/bsd/xfs_vnodeops-netbsd.c (xfs_netbsd_{read,write}): return
        EISDIR for VDIR, return EFBIG for negative offsets, update
        read/write creds

        * arlad/messages.c (possibly_have_network): new function, return
        non zero if there possibly is network connectivity
        (viocconnect): when going to disconnectivity mode, check for if there
        is network with possibly_have_network() before giving up callbacks
        
        * arlad/volcache.c (volcache_getby{name,id}): if the volume
        changed (stablep == FALSE), then remove all markings since they
        will not be correct now

        * arlad/fcache.h: remove fs_server_context.conns[].status, it
        should never have been there and its wasn't used either.
        
        * arlad/poller.[ch]: new file

        * arlad/Makefile.in: common_SRCS,common_OBJS += poller.[co]
        
        * arlad/arla.c (arla_init): init poller
        
        * arlad/arla_local.h: move up hash.h, add heap.h and poller.h
        
        * arlad/volcache.c (volcache_timeout): volcache entry timeout
        (recycle_entry): clear status
        (volcache_process_marks): process mark on the connection
        (volume_uptodatep): volume not uptodate of the volume is too old
        (get_info_common): set timeout time
        (get_info_loop): change signature to allow passing in a nvldbentry to
        be update instead of e->entry.
        (get_info_byid): change signature to fit get_info_loop
        (get_info_byname): change signature to fit get_info_loop, wake up any
        waiters when done.
        (volcache_getbyname): instead of passing in a VolCacheEntry, use a
        nvldbentry when comparing the entry with the data to determin if
        its stable.
        (volcache_getbyid): ditto
        (volume_downp): test if it was a volume error
        (volcache_mark_down): mark the site down appropriately
        (volcache_reliablep_el): return true if the site is a good one
        (volcache_reliable_el): set that the site is a good one
        (volcache_reliablep): rename from volcache_reliable
        (volume_make_uptodate): changed signature of get_info_byname

        * arlad/volcache.h (volcacheentry): add timeout and status entry,
        remove last_fetch add volume server/partion fags
        (volume_downp): new function
        (void volcache_mark_down): ditto
        (void volcache_process_marks): ditto
        (volcache_reliablep_el): ditto
        (volcache_reliable_el): ditto
        (volcache_reliablep): old volcache_reliable
        (VOLCACHE_TIMEOUT): timeout

        * arlad/messages.c: s/volcache_reliable/volcache_reliablep/
        (setrxkcrypt): only clear rxkad conns

        * arlad/fcache.c (collectstats_stop): simplify and use data that
        now in connection and volcache entry (passes in)
        (fs_probe): export
        (fcache_poller_unref): new function
        (fcache_poller_reref): new function
        (throw_entry): remove ref to poller
        (create_nodes): init entry->poll to NULL
        (invalidator): remove ref to poller
        (fs_rtt_cmp): new function, compare to pointers to two fs_server_entry.
        (find_partition): find partition of the context that we used
        (find_next_fs): rework, now an error is passed in and the old
        entry/site is marked dead if there was such an error, if there was
        no error with last volume, mark it good. Modify to changes of the
        fs_server_context.
        (free_fs_server_context): Modify to changes of the fs_server_context,
        also call volcache_process_marks().
        (find_first_fs): insert a middle case, between the server is bad
        (INT_MAX) and the server it up (smoth rtt) there is volume bad
        (INT_MAX / 2). Modify to changes of the fs_server_context.
        (stale): remove ref to poller
        (update_entry): pass in a conn instead of a ip-address, it is to make
        it eaiser to find the real identity of the server we talked to.
        ref the poller if we have connection, and unref the poller in
        there was none (disconnected)
        (update_attr_entry): pass in a conn instead of a ip-address.
        (resolve_mp): fetch the whole symlink
        (*): deal with that update_entry, update_attr_entry, find_next_fs
        use conn_ref the make it possible to send conn's around.
        
        * arlad/fcache.h (FCacheEntry): add poller pointer
        (fs_server_context): add status and entry number for a conns[N]
        (fs_probe): export function

        * arlad/conn.[ch] (conn_ref): new function
        
        * arlad/dump_state.c (prefix): use ARLACACHEDIR
        (print_fcache_entry): print the cache node index

2002-05-12  Love  <lha@stacken.kth.se>

        * arlad/cmcb.c: update (c)
        
        * appl/mpp/mpp.c (main): use growing hashtable
        
        * util/hash.[ch]: growing hashtable support, api compatible

        * util/arlamath.[ch],Makefile.in,util-tester.c: prime-number testing
        functions, by mattiasa@e.kth.se, testing by me

        * arlad/fcache.c (new_fetched_length): common code that caculates
        the new fetched length
        (*): use new_fetched_length

2002-05-10  Love  <lha@stacken.kth.se>

        * rx/rx_rdwr.c: turn a #ifndef to a #ifdef (like on the other
        places ADAPT_WINDOW is used) make the code somewhat easier to
        follow

        * lib/ko/koerror.c: move #includes to avoid `MAXKTCTICKETLEN'
        redefined problem

2002-05-07  Love  <lha@stacken.kth.se>

        * appl/mpp/mpp.c (macro_define): make it handle value == NULL
        (process_define): when there is no value, pass in NULL

        * tests/*: make !-fast test less verbose when running in -fast mode

        * lib/cmd/Makefile.in: use DESTDIR when installing man page
        
        * arlad/fcache.c (trucate_file): mirror code from write_data, now
        files don't get full of NULL when using program that ftruncate()
        files before they close() them (like rcs and install)
        
        * doc/themis.texi: add common things to exclude and default
        exclude items

        * arlad/fcache.c: (write_data): If the hole file is fetched as we
        last saw it, lets write down the whole file to the fileserver. If
        the file is shrinking, make sure we don't cache non-existing
        bytes.
        
        This fixes problems with previous commit that I should have
        noticed

        * tests/part-files: make test smaller/faster, add some comments
        
        * tests/run-tests.in (ARLA_TESTS): add part-files
        
        * tests/part-files: test if files are written/read correctly when
        they are partly modified

        * arlad/fcache.c (write_data): only write the content of the file
        we have fetched

2002-05-06  Love  <lha@stacken.kth.se>

        * arlad/arla.c (conf_params): trim it to make newer gcc happy
        
        * xfs/linux/xfs_inodeops.c (xfs_release_file): only look at the
        I_DIRTY_DATASYNC|I_DIRTY_PAGES flags, not at all flags I_DIRTY
        since that is also set for meta data changes (atime modifcation)
        
        * util/util-tester.c (test_hash): hashtab's are freed by
        hashtabrelease()

        * appl/mpp/mpp.c: add macro undefine (-U and %undef)
        
        * arlad/arla.conf.5, conf/arla.conf: document sysname
        
        * arlad/fcache.[ch],adir.c: We need to keep track of how much of
        file is cached so when the truncate is done we wont use and NUL
        after the cache file

        XXX optimization: if we have the whole file and we are truncating
        to a larger will, we already have the data, so we should update
        fetched_length to the new size.

2002-05-05  Love  <lha@stacken.kth.se>

        * configure.in (AC_OUTPUT): added appl/mpp/Makefile
        
        * appl/Makefile.in (SUBDIRS): add mpp
        
        * appl/mpp/mpp.1: manpage
        
        * appl/mpp/mpp.c: add command line parsing, fix space (' ')
        problems

2002-04-30  Love  <lha@stacken.kth.se>

        * appl/mpp: add a almost compatible mpp processor, memory hungry
        and with a bad attitude toward space

        * arlad/fcache.c (collectstats_getentry): replace some 32 that
        really was HISTOGRAM_SLOTS

        * arlad/messages.c (aioc_statistics): use constants, rewrite if
        ().. to switch()

2002-04-29  Love  <lha@stacken.kth.se>

        * arlad/arla.c (read_conffile): enable parser to parse strings too
        (*): keep up with above change
        init varibles
        
        * arlad/arla-cli.c: renamed the holder of the sysname passed in
        vid the argv (and getarg()) to argv_sysname, arla_init changed
        signature

        * arlad/arlad.c: renamed the holder of the sysname passed in vid
        the argv (and getarg()) to argv_sysname, arla_init changed
        signature

        * arlad/arla_local.h: arla_init changed signature, renamed the
        holder of the sysname passed in vid the argv (and getarg()) to
        argv_sysname

        * arlad/messages.c (vioc_afs_sysname): rename the variable that
        hold the temporary sysname

        * arlad/fcache.c (fprioritylevel): set default level here
        
        * appl/vos/vos_examine.c (print_extended_stats): when in extented
        mode, print out filecount too
        (print_volume): change where \n are added

        * doc/pioctl.texi: arla pioctl
        
        * configure.in: AM_CONDITIONAL's that roken will check (and
        disable if necessary) before roken fragment
        
        * configure.in: reverse the libtool vs automake stuff
        
        * HACKING: depend on libtool too
        
        * appl/cbdebug/Makefile.in: sl depend on readline
        
        * configure.in: move down the kerberos tests to after we check for
        libroken

        * cf/check-roken.m4: when using an external roken, use -L$dir
        -lroken, its not sure that there exists a .a file
        
        * cf/check-kerberos.m4: check for system kerberos 5 headers first,
        this mirrors the pattern of searching for a lib

2002-04-29  Mattias Amnefelt  <mattiasa@e.kth.se>

        * xfs/bsd/xfs_vnodeops-osf.c: (1.23.2.4 -> 1.23.2.5) (xfs_fsync) do
        ubc_flush_dirty even if !(fflags & FSYNC)
        * xfs/bsd/xfs_vnodeops-osf.c: (xfs_fsync): Flush dirty ubc buffers

2002-04-29  Love  <lha@stacken.kth.se>

        * ydr/output.c (encode_varray): varray's w/o limit is of size 0,
        not -1

        * arlad/messages.c: sprinkel some all_powerful_p()
        
2002-04-28  Love  <lha@stacken.kth.se>

        * rx/{rx_dumptrace.c,rx_trace.[ch]}: these are covered by IBM (c)
        
        * rx/Makefile.in: remove hack to build rx_trace
        
        * rx/rx_trace.h: add stuff that is shared between rx_trace.c and
        rx_dumptrace.c

        * rx/rx_trace.c: remove stuff that is shared with rx_dumptrace.c
        
        * rx/rx_dumptrace.c: debug program part of rx_trace.c
        
        * rx/rx_knet.c: remove, not part of arla rx but rather openafs rx

        * rx/rx_user.c (rxi_GetUDPSocket,rxi_Listener): keep track of all
        sockets we open rxi_Listener can only probe those sockets that are
        are rx's instead of iterating over all sockets 0 ->
        rx_maxSocketNumber. This matters for win32 enviroment where
        sockets typically are large numbers.
        
        (rxi_Listener): always use IOMGR_Select instead of doing select in
        case we are doing a poll since that is was IOMGR select will do
        for us if we send in a "poll-timeval" (sec = usec = 0). This fixes
        the problem found by Tomas Olsson that select() (the system
        version) might have a diffrent calling convention then
        IOMGR_Select(), lets say PASCAL. Casting system() and
        IOMGR_Select() to the same function pointer will have have fatal
        problems.
        
        Also when here, clean up the code some (like use NULL, etc)

2002-04-27  Love  <lha@stacken.kth.se>

        * TODO: some fixes, some more comments
        
        * HACKING: update what version you need

        * xfs/bsd/**[ch]: split xfs_node->cred into rd_cred and wr_cred
        and use appropriately

        * acconfig.h: remove, now autoconf/autoheader does this

        * configure.in: pacify autoconf 2.53
        
        use AC_HAVE_TYPES instead of AC_GROK_TYPES
        
        do libtool vs automake stuff
        
        * cf/kernel*: use patsubst

        * __extentions__.m4: __EXTENSIONS__ hack

        * cf/misc.m4: fool autoconf 2.53 by using \x40 instead of \100
        
        * xfs/bsd/bin/mount_xfs.c: start use setprogname()
        
        * arlad/messages.c: cred_list_pag changed signature
        
        * arlad/cred.h: cred_list_pag changed signature
        
        * arlad/cred.c: make the cred list be on the pair `cred,type' to
        make diffrence between CRED_KRB4 and CRED_NONE.
        
        (cred_free): ignore CRED_ROOT_CELL entries, they will be g/c when the
        last referrer releases it.
        
        also make code more consistent.
        
2002-04-26  Love  <lha@stacken.kth.se>

        * include/kafs.h: allocate AIOC_PTSNAMETOID
        
        * rxdef/ka.xg: protect MAXKTCTICKETLEN
        
2002-04-26  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * configure.in: Added -traditional-cpp to CPPFLAGS on Darwin

        * Makefile.in: Added CPPFLAGS to ydr's CPP variable

2002-04-26  Love  <lha@stacken.kth.se>

        * ydr/output.c (generate_hdr_enum): print out the enum too
        * lib/roken/roken.h: now generated

        * ydr/output.c: fix some more don't iterate on empty lists
        
        * ydr/output.c: structs w/o members are 0 sized.

        * appl/cbdebug/cbdebug.1: new manpage
        
        * **/*.c: use setprogname() and getprogname()

        * configure.in: check for <glob.h> and <vis.h> and so
        AM_CONDITIONAL stuff

        * cf/check-kerberos.m4: always run AM_CONDITIONAL(KRB{4,5})

        * xfs/linux/xfs_dev.c: indent
        
2002-04-25  Love  <lha@stacken.kth.se>

        * appl/cbdebug/cbdebug.h,Makefile.in: added real parsing and broke
        out whoareyou code.
        
        * appl/cbdebug/Makefile: some fixes
        
        * appl/cbdebug/whoareyou.c: some fixes
        
        * arlad/cmcb.c: some fixes, 
        (init_address): don't fail if there isn't any address

        * appl/cbdebug/cbdebug.c: move out all useful things, add sl
        command parsing

        * appl/cbdebug/whoareyou.c (print_addr): move here
        (whoareyou): add agetarg command parsing

        * arlad/cmcb.c (init_address): break out the initialization
        interfaceAddr
        (RXAFSCB_InitCallBackState2): use init_address()
        (RXAFSCB_WhoAreYou): use init_address()

2002-04-24  Love  <lha@stacken.kth.se>

        * NEWS: added UUID
        
        * appl/Makefile.in (SUBDIRS): add cbdebug
        
        * include/Makefile.in: add ifaddrs.h and rtbl.h
        
        * configure.in: check for <ifaddrs.h> and provide automake glue
        add appl/cbdebug
        
        * lib/ko/afs_uuid.[ch]: const poison compare functions

        * arlad/arla_local.h: added <ifaddrs.h>
        
        * arlad/cmcb.c: implement RXAFSCB_ProbeUUID, RXAFSCB_WhoAreYou,
        RXAFSCB_InitCallBackState2, and, RXAFSCB_InitCallBackState3
        
        InitCallBackState{2,3} functions are implemented by simply calling
        the common part of InitCallBackState
        
        Move the InitCallBackState functions next to each other.
        
        XXX Need to fix the mtu detection and netmask-less ifa's
        * appl/cbdebug/cbdebug.c: add <afs_uuid.h>
        
        * appl/cbdebug: a simple callback manager debug program
        
        * include/Makefile.in: added afs_uuid.h
        
        * arlad/arla_local.h: "afs_uuid.h"
        
        * lib/ko/Makefile.in: LIB_{SOURCES,OBJECTS} += afs_uuid.[ch]
        
        * lib/ko/afs_uuid.c: fix comment about MS GUID
        
        * lib/ko/afs_uuid.c: protect the headers <net/if.h>,
        <net/if_dl.h>, and <net/if_types.h>

        * configure.in: check after <net/if.h>, <net/if_dl.h>, and
        <net/if_types.h>

        * lib/ko/Makefile.in: sort LIB_OBJECTS and LIB_SOURCES
        
        * have-struct-field.m4: fix another autoconf 2.52 fallout
        
        * cf/have-linux-kernel-types.m4: make it work with autoconf 2.52

        * cf/have-linux-kernel-type.m4: make it work with autoconf 2.52's
        AC_MSG_RESULT

        * HACKING: update what version you need, enable users to pass in
        ACLOCAL_EXTRA_ARGS as enviroment variable
        
2002-04-23  Love  <lha@stacken.kth.se>

        * **/Makefile.in: use libtool
        
        * arlad/arla-cli.c,arlad.c,dump_state.c: use <version.h> again

        * configure.in, acconfig.h, missing:
        update to 2.52, use libtool and update to

        * appl/afsmgr/Makefile.in: remove udebug, its not built here
        
        * lib/roken/*: merge heimdal 2002-04-23 code
        
        * cf/*: merge heimdal 2002-04-22 code

2002-04-23  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/mac/arlacmm/permwin.c:
        Add beep when apply fails. Make Apply button inactive when not needed

        * appl/mac/arlacmm/arlacmm.h: Added aclcopy to ArlaCMMType

        * appl/mac/arlacmm/dialog.nib/objects.xib,
        appl/mac/arlacmm/dialog.nib/info.nib: adduser: make ok button
        default

        * appl/mac/arlacmm/permwin.c (addAclItem): added duplicate check

        * appl/mac/arlacmm/arlacmm.h: Added cfstring_to_utf8 and verifyname

        * appl/mac/arlacmm/util.c: Added cfstring_to_utf8

        * appl/mac/arlacmm/adduserwin.c, appl/mac/arlacmm/permwin.c,
        appl/mac/arlacmm/arlacmm.h: Add user

        * appl/mac/arlacmm/permwin.c: Added remove command

2002-04-23  Love  <lha@stacken.kth.se>

        * arlad/{fcache,messages}.c,arla_local.h:
        hack to disable kernelp asserts, the bug/feature of xfs should
        really be fixed but I don't have a clue to how to fix it now

2002-04-23  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/mac/arlacmm/main.c, appl/mac/arlacmm/arlacmm.h,
        appl/mac/arlacmm/pioctl.c: Only show permissions in the contextual
        menu if it is a directory

        * appl/mac/arlacmm/dialog.nib/objects.xib,
        appl/mac/arlacmm/dialog.nib/info.nib, appl/mac/arlacmm/pioctl.c,
        appl/mac/arlacmm/main.c, appl/mac/arlacmm/permwin.c,
        appl/mac/arlacmm/arlacmm.h, appl/mac/arlacmm/infowin.c,
        appl/mac/arlacmm/Makefile.in, appl/mac/arlacmm/adduserwin.c: Added
        info window and fix CFRelease bug

        * appl/mac/arlacmm/permwin.c: Added an event handler for window
        closing

        * appl/mac/arlacmm/permwin.c (applyAcl): Added

        * appl/mac/arlacmm/pioctl.c: Added setacl

        * appl/mac/arlacmm/arlacmm.h: Added path and setacl

        * appl/mac/arlacmm/Makefile.in: Use install instead of tar

        * appl/mac/arlacmm/permwin.c (changeAclItem): added
        
        * appl/mac/arlacmm/adduserwin.c (fillAddUserGroupsList): group
        array is immutable
        
        * appl/mac/arlacmm/arlacmm.h: acl and group arrays are mutable

2002-04-21  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * fs_getprio.c (getprio_cmd): correct arg usage

        * fs.c (setprio_cmd): correct arg usage
        (getmaxprio_cmd): correct arg usage

        * conn.c (pinger): ignore hosts in disconnected mode

        * fcache.c (fcache_reobtain_callbacks): use cred
        (fcache_get): check if volume is cached before creating entry
        (fcache_verify_attr): get all rights when disconnected

        * fcache.h: fcache_reobtain_callbacks takes cred

        * messages.c:
        Always set XFS_ATTR_R in fcacheentry before installing attr
        (viocconnect): call fcache_reobtain_callbacks with creds

        * volcache.c (get_info_common): don't return -1
        (volcache_getbyname): return ENETDOWN if disconnected and volume
        isn't cached

2002-04-21  Love  <lha@stacken.kth.se>

        * xfs/bsd/xfs_syscalls-common.c: do the XFS_NOT_LKM right for
        xfs_setgroups

        * xfs/bsd/xfs_syscalls-common.c: XFS_NOT_LKM: symbol that tells us
        that we are no an lkm/kld/...

        * xfs/bsd/xfs_wrap-bsd.c (freebsd): name the module arlaxfsdev
        instead of xfsdev

        * xfs/bsd/xfs_wrap-bsd.c: freebsd don't support LKM, remove ifdef
        
        * xfs/bsd/xfs_vfsops-freebsd.c (xfs_vget_freebsd): provide now
        that freebsd have 4 argument (extra argument, locking flags), and
        use it.

        * cf/check-kernel.m4: use to nm trick for freebsd5 too
        
        * xfs/bsd/xfs_syscalls-common.c: Create a xfs_crcopy that works
        around the problem that around freebsd version 500026, crcopy
        changed semantics.

        * xfs/bsd/xfs_common-bsd.c (xfs_suser): freebsd current have
        changed from struct proc * to struct thread * as argument to
        suser() in current

        * xfs/bsd/xfs/xfs_locl.h: move those things that depend on
        HAVE_FREEBSD_THREAD to after we define it

        * lib/ko/afs_uuid.c (afsUUID_from_string): implement
        
2002-04-20  Love  <lha@stacken.kth.se>

        * lib/ko/afs_uuid.[ch]: ncs/dce/afs/guid generator

        * arlad/messages.c (gettok_func): new healper function to
        viocgettok()
        (viocgettok): use the new cred api to iterate over the creds
        this brings the cpu usage of arla down from 50-60% to 4% on my
        laptop (with 5 tokens) when doing:
        
           while :; do tokens >/dev/null ; done

        * arlad/cred.c: All creds in the same pag (xfs_cred_t) are now
        stored on a linked list to make it possible to iterate over all
        creds in the same pag fast.
        
        The head is inserted in the hash, use the magic cellid == -1 to
        mark a head.
        
        (list_pag_func): helper function, iterate over all creds in same pag
        (cred_list_pag): iterator function that lets user iterate over all
        cred in same pag
        (recycle_entry): if not root entry, remove it from the list
        (cred_free): assert its not a root entry
        (root_free): remove the entry from the pag-list and if the list is
        empty, free the root-head.
        (add_to_root_entry): add the entry to the root head, create the
        root head if it doesn't exist.
        
        
        * arlad/cred.h (CredCacheEntry): add the pag union
        (cred_list_pag): new function

        * util/list.c: implement listfree()
        
        * util/list.h: add listfree()
        
        * appl/kalog/kalog.c appl/lib/appl_locl.h appl/lib/ka-procs.c
        arlad/arla_local.h arlad/cred.h lib/ko/koerror.c
        milko/appl/perf/perf.c milko/bos/bos_locl.h milko/fs/fsrv_locl.h
        milko/lib/msecurity/msecurity.c milko/lib/msecurity/netinit.c
        milko/pts/pr.c milko/pts/ptserver.c milko/vldb/vldb_locl.h
        rxkad/rxkad_locl.h: if using openssl, include <openssl/des.h>
        instead of <des.h>, work for everything using openssl except Mac
        OS X, oh joy

2002-04-18  Love  <lha@stacken.kth.se>

        * appl/fs/fs_statistics.c (getstatistics_cmd): host is not a type
        
2002-04-17  Love  <lha@stacken.kth.se>

        * arlad/fcache.c (stale): only so throw_data if there is data
        (flags->datap)

        * arlad/fcache.c (stale): if its a directory, lets throw the data
        directly, this mirrors the behavior in the kernel
        
        * arlad/messages.c: change signature of getvolstat
        
        * arlad/fcache.c: change signature of getvolstat, now pass length
        of data, this make sure we know the length of the target string
        and can use it (and not the size of the pointer)
        
        * arlad/fcache.h: change signature of getvolstat
        
        * rxdef/cb.xg: %#include <fs.h>
        * rxdef/vldb.xg: %#include <fs.h>
        
        * rxdef/common.h: remove the definition of struct afsUUID and
        instead include "afsuuid.h"
        
        * rxdef/Makefile.in: make vldb and cb files depend on fs.h
        
        * rxdef/fs.xg: define AFSUUID_GENERATE
        
2002-04-16  Love  <lha@stacken.kth.se>

        * rxdef/afsuuid.h: struct afsUUID;
        
        * arlad/dynroot.c: remove the local variable dynrootcell and start
        to use DYNROOT_CELLID

        * lib/ko/ko.h: export DYNROOT_CELLID
        
        * lib/ko/kocell.c (update_cell): dynroot cell is uptodate, this
        stops zillions of dns lookups

        * appl/vos/vos_listvldb.c: be compatible with transarc/openafs,
        its not ... Site %s, but %s Site

2002-04-15  Love  <lha@stacken.kth.se>

        * ydr/output.c: do bounce checking of TVARRAY types
        s/bcopy/memcpy/
        
        * rx/rx.c (rx_NewConnection): if RXS_NewConnection fails, set its
        error to conn->error, not really right, but then the interface
        isn't that good
        (rx_NewService): when creating a new service call RXS_NewService()
        
        * rx/rx.h: define RXS_NewService and use appropriate element in
        rx_securityClass

        * Checking in lex.l, parse.y, sym.h, symbol.c, types.c, types.h:
        add support for __attribute__ on structs, specifically __nogenerate__

        * ydr/output.c: generate the max-size for definitions for limited
        dynamic types too

2002-04-12  Love  <lha@stacken.kth.se>

        * appl/fs/fs.c (cmds): give "sp" to setserverprefs since
        setpriority isn't used (and never has been used), compat with
        openafs/transarc afs, ident

        * tests/blocks-new-file.c: make the file smaller (64K instead of
        1M), this makes the test fly by when running on a slow link
        
        * arlad/{messages.c,fcache.c,inter.c}:

        Change fcache_verify_data() to fcache_get_data() on users that
        will use the data later, now all callers don't need to first
        fcache_verify_attr() and then get ->wanted_length since
        fcache_get_data() will do that for them.

        Due to the signature diffrence of fcache_verify_data() and
        fcache_get_data() there is some prototype changes on other
        functions to deal with this.

        * arlad/messages.c (read_mount_point): update wanted_length so we
        don't fall over when doing a fs lsm on a (data) uncached entry
        
2002-04-10  Johan Danielsson  <joda@pdc.kth.se>

        * conf/arla.conf: spelling
        
        * appl/fs/fs.c: add some more command aliases

2002-04-10  Love  <lha@stacken.kth.se>

        * arlad/*.[ch]: multi-sysname implementation
        
        * arlad/.gdbinit: print out the thread index too in lwp_ps

        * arlad/messages.c (xfs_message_inactivenode): comment out the
        kernelp assert for now

        * xfs/linux/xfs_inodeops.c (xfs_mightwrite_p): new function that
        checks if a vma will write to disk
        (*): use xfs_mightwrite_p

2002-04-09  Love  <lha@stacken.kth.se>

        * scripts/mingwin.sh: bootstrap arla with mingw
        
2002-04-08  Love  <lha@stacken.kth.se>

        * xfs/include/xfs/xfs_messages.h,xfs/bsd/xfs_messages.c,
        xfs/bsd/xfs_vnodeops-netbsd.c: use temporary flag XFS_VMOPEN when
        the node is locked so installdata/attr wont touch the vnode,
        should stop useing genfs instead

        * arlad/arlad.c: update copyright
        
        * xfs/bsd/Makefile.in: generate function prototypes for the VOPS
        
        * xfs/bsd/xfs_node-bsd.c: init genfs node for netbsd
        
        * xfs/bsd/xfs_vfsops-netbsd.c: init the vnodeopv_desc depending on
        if we are using ubc or not, (xfs_mount_netbsd): init mnt_*_bshift
        for genfs
        
        * xfs/bsd/xfs_vnodeops-bsd.c: make all xfs vops global
        (xfs_pushdirty): make to work for netbsd UBC
        
        * xfs/bsd/xfs_vnodeops-common.c: (xfs_open_common): make it always
        store the cred, it will be needed
        (xfs_close_common): push dirty block in case of a VREG
        (xfs_uio_end_length): rename uio_end_length from export and optimize
        (*): use changed signature of xfs_uio_end_length
        
        * xfs/bsd/xfs/xfs_locl.h: move xfs_pushdirty() to later, include
        genfs stuff for netbsd ubc code, netbsd have removed UVM define,
        provide a macro for XFS_VOP_DEF that creates a prototype for all VOPs

        * xfs/bsd/xfs_vnodeops-netbsd.c: netbsd UBC code

        * xfs/bsd/xfs/xfs_node.h: include genfs for netbsd ubc
        * xfs/bsd/xfs/xfs_vnodeops.h: add prototype for xfs_uio_end_length

2002-04-02  Love  <lha@stacken.kth.se>

        * appl/amon/Makefile.in: set LDFLAGS to @LDFLAGS@ so we use it
        
        * xfs/linux/xfs_inodeops.c: change all occasions of VM_MAYWRITE to
        VM_WRITE since linux maps binaries on i383 with
        VM_MAPWRITE. VM_MAY flags are "maximum rights" for mprotect as I
        understands it.

2002-03-22  Love  <lha@stacken.kth.se>

        * configure.in: openbsd might need CPPFLAGS too for kernel flags
        test in the future, give to them
        
2002-03-19  Tomas Olsson  <tol@stacken.kth.se>

        * xfs/winnt/bin/wakeup.c: added
        * xfs/winnt/bin/makefile: add wakeup, clean a bit
        * xfs/winnt/bin/xfs-helper.c: make it work
        * xfs/winnt/{inc,src}/*: read (sort of)
        * lib/bufdir/fdir.c (first_slotp): add & use
        * lib/bufdir/afs_dir.h: added AFSDIR_FIRST

2002-03-14  Love  <lha@stacken.kth.se>

        * xfs/bsd/xfs_vnodeops-common.c (xfs_fsync_common): replace
        ubc_pushdirty() with xfs_pushdirty()

        * xfs/bsd/xfs_vnodeops-bsd.c (xfs_pushdirty): flush dirty blocks
        to disk

        * xfs/bsd/xfs_vnodeops-bsd.c (xfs_putpages): netbsd-ubc case, only
        mark VREG files dirty if they are

2002-03-12  Love  <lha@stacken.kth.se>

        * xfs/bsd/xfs_vnodeops-bsd.c (xfs_putpages): only mark node as
        dirty if there are dirty pages in the cache node in the
        netbsd-current case.
        
        XXX needs to be fixed a better way

        * tests/asu.c: If the test is ran w/o pag, we can't switch user
        since that will mean we will loose credentials.
        
2002-03-11  Love  <lha@stacken.kth.se>

        * xfs/bsd/xfs_vnodeops-bsd.c (xfs_putpages): When a directory node
        is reclaimed because of a xfs_message_invalidnode, vinvalbuf is
        called which triggers a VOP_PUTPAGES, but then there is no data
        the underlaing VOP_PUTPAGES can't be called.
        
        So, if you want to avoid getting (bogus) tar errors (when run as
        root) with creds, you better run in a pag.
        
2002-03-08  Love  <lha@stacken.kth.se>

        * arlad/arla-cli.c: output formating error, from kolya@mit.edu
        * appl/fs/fs_diskfree.c: formating errors, from kolya@mit.edu

2002-03-07  Love  <lha@stacken.kth.se>

        * arlad/fcache.c: Make (cm_f)truncate(_file) take a storestatus
        argument to avoid a race with changing permissions, modify the
        filesize after the rpc completes with nonerror. Bug noticed by
        kolya@mit.edu.
        * arlad/fcache.h: Ditto
        * arlad/inter.c: Ditto
        * arlad/inter.h: Ditto
        * arlad/messages.c: Ditto
        * arlad/reconnect.c: Ditto

2002-03-06  Tomas Olsson  <tol@stacken.kth.se>

        * arlad/messages.c (xfs_message_create): use fcache_conv_file_name
        * arlad/inter.c (cm_open): use fcache_conv_file_name
        * arlad/arla-cli.c (print_dir{,long}): return int
        * arlad/fcache.c (fcache_conv_file_name): added
        (fcache_fh{open,get}): adjust for cygwin
        (bulkstat_help_func): return int

        * arlad/fcache.h: added fcache_conv_file_name()
        * arlad/adir.h (adir_readdir): use new fdir_readdir_func
        * arlad/adir.c (adir_readdir): call fdir_readdir the new way
        * arlad/kernel.c (tcp_open): use the right address, return socket
        (tcp_write): error checking

        * arlad/unknown-subr.c: adjust for cygwin/nt
        * include/Makefile.in: added fids.h
        * lib/ko/kocell.c (fetch_host,dns_lookup_cell): don't get RR if we
        don't know how
        * lib/bufdir/fbuf.h: ifdef for kernel
        * lib/bufdir/fdir.h (fdir_readdir_func): return int (time to stop?)
        (fdir_readdir): accept offset

        * lib/bufdir/fdir.c (fdir_readdir): accept & maintain offset make
        kernel safe
        * arla/lib/bufdir/fids.h added, contains AFSfid & VenusFid
        
2002-03-04  Jimmy Engelbrecht  <jimmy@heliga-birgitta.stacken.kth.se>

        * arlad/messages.c, include/kafs.h, appl/lib/fs_lib.c:
        Merged in VIOC to AIOC interface change for [VA]IOC_STATISTICS from
        0.35-branch

2002-03-04  Mattias Amnefelt  <mattiasa@heliga-birgitta.stacken.kth.se>

        * arla_local.h: rx/rxgencon.h is included in rx.h

2002-03-04  Mattias Amnefelt  <mattiasa@e.kth.se>

        * xfs/linux/xfs/xfs_locl.h: made xfs_dcount() arg const
        * arlad/arla_local.h: rx/rxgencon.h is included in rx.h

2002-03-02  Love  <lha@stacken.kth.se>

        * ydr/lex.l: handle afs_int32 and afs_uint32

2002-02-27  Love  <lha@stacken.kth.se>

        * conf/CellServDB: add dementia.org and grand.central.org

        * util/mmaptime{,_test}.c: make it work when there is not working mmap

        * rx/Makefile.in: don't link rxdebug or rx_trace with libcmd since
        it doesn't use it

        * configure.in: check for winsock.h, add case for mingw
        
        * appl/lib/arlalib.c: (key_to_key): remove unused function,
        (arlalib_get_cred_krb): kill some const warnings by coping the
        string
        
        * lib/cmd/Makefile.in: don't install testc_1_MAN

2002-02-25  Love  <lha@stacken.kth.se>

        * Release 0.35.7, see the branch for relevant patches

        * scripts/run-tests.sh: release testing
        * scripts/extract-result.sh: extract test info from run-tests
        * scripts/build-arla.sh: release build test script
        
2002-02-24  Love  <lha@stacken.kth.se>

        * tests/fhbench.c: 1.6->1.6.2.1: include timevalsub and rename it
        to tvsub

        * xfs/linux/xfs_{common,message}.c: (print_aliases): print out if
        the name is on (name)hash or not
        (users of ->d_count): use helper function xfs_dcount

2002-02-24  Mattias Amnefelt  <mattiasa@e.kth.se>

        * tools/gnats/Makefile.in: install into DESTDIR

2002-02-18  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * doc/themis.texi: Document the M directive and the P flag.

2002-02-08  Love  <lha@stacken.kth.se>

        * xfs/linux/xfs_inodeops.c (xfs_file_mmap): pass in the xfs-inode
        to data_valid instead of the cache-node

2002-02-07  Love  <lha@stacken.kth.se>

        * All files:
        
        s/u_int17_t/uint17_t/
        s/libutil.a/libarlautil.a/
        remove things from arla's libroken that doesn't belong there
         (that is, it doesn't exists in other-libroken), and move
         then to arlautil
        make compile on netbsd-current (CPPFLAGS i needed, more <sys/time.h>)
        remove things that isn't needed in <xfs/xfs_message.h>
         to operating system depended headers/arlad headers
         this is to make it easier to use code in kernel
        
       * configure.in:  how the roken thingy work now, you have
        to specify --with-roken if there you are going to link agaist a
        roken in the kerberos test. This is due to that we now link agaist
        the .a lib directly.

2002-02-06  Love  <lha@stacken.kth.se>

        * xfs/bsd/xfs_syscalls-common.c (remote_pioctl): not that it is
        used, set the message size in the message header to unconfuse
        readers

        * xfs/linux/xfs_dev.c (free_node): create yet another free
        function, this one does consistency checks and replaces
        xfs_free(xn, XFS_MEM_XNODE)
        (xfs_queue_inactive): if the channel is closed, just free the node
        (xfs_process_inactive_queue): use free_node()
        (xfs_empty_inactive_queue): use free_node()

        * xfs/linux/xfs_syscalls.c (xfs_fh_to_dentry): linux-2.4: after a
        get_super, we need to drop_super(), w/o this, the we will fail to
        unmount the super block of the cache-filesystem.
        
2002-02-04  Love  <lha@stacken.kth.se>

        * ydr/*.[chly]: rename Tfoo symbols to YDR_Tfoo to avoid clashes
        with toolchains with TCHAR

2002-02-01  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/mac/Arla_Configuration/ReadCells.m: Added static on local
        functions
        * appl/mac/Arla_Configuration/Makefile.in: Use install instead of
        tar
        * appl/mac/Arla_Configuration/DataSource.m: Added static on local
        functions
        (checkdaemonpid): give -p to ps

2002-02-01  Love  <lha@stacken.kth.se>

        * cf/check-roken.m4: 1.1->1.1.2.1: Reference libroken directly
        instead of with -l

        * lwp/Makefile.in: 1.58->1.58.2.1: Remove superfluous -lroken
        
        * xfs/linux: dual license xfs for linux under bsd/gpl, update
        copyright years. See xfs/linux/license for more detailed
        information.
        
        * conf/CellServDB (theory.cornell.edu): removed, from Ruth
        Mitchell <mitchell@tc.cornell.edu> on info-afs@transarc.com
        
        * xfs/linux/xfs_dev.c (xfs_process_inactive_queue): make sure we
        don't add node twice if we race with someone to
        xfs_queue_inactive()

        * xfs/linux/xfs_message.c: minimize diff to 0.35-branch
        
2002-01-31  Love  <lha@stacken.kth.se>

        * arlad/arlad.c (arla_start): after chroot(), do a chdir("/")
        
        * cf/linux-func-init-mutex.m4: make test work when init_MUTEX is a
        macro (late linux-2.2)

2002-01-27  Love  <lha@stacken.kth.se>

        * scripts/arla-release.sh: spelling
        
2002-01-26  Love  <lha@stacken.kth.se>

        * conf/CellServDB: added su.se
        
        * tests/run-tests.in: add new option -fail that make test-run stop
        after the first failure

        * xfs/linux/xfs_inodeops.c: (xfs_d_delete): use
        xfs_queue_inactive() again since it correct

        * xfs/linux/xfs_message.c (xfs_invalid_xnode): use
        xfs_queue_inactive() again since it correct

        * xfs/linux/xfs_dev.c (xfs_queue_inactive): make sure we don't
        enter an node twice on the inactive_list
        (xfs_process_inactive_queue): use list_del_init so the inactive_list
        have known state
        (xfs_empty_inactive_queue): don't free nodes that have shouldn't
        be freed

        * xfs/linux/xfs_vfsops.c (print_nodes): use new node_list list
        macro and protect
        (xfs_read_super): init the node_list and node_sem fields

        * xfs/linux/xfs_node.c (xfs_iget): move the null address space ops
        here, and update the code
        (xfs_iget): use new node_list list macro and protect
        (free_xfs_node): ditto
        (free_all_xfs_nodes): ditto
        (xfs_node_find): ditto

        * xfs/linux/xfs_message.c (xfs_invalid_xnode): use
        xfs_message_send to notify userland that node is inactive,
        xfs_queue_inactive() have unwanted side-effects
        (xfs_message_gc_nodes): remove the case where we gc all nodes. its not
        used for arlad, and locking became horrible to solve.

        * xfs/linux/xfs_inodeops.c:
        o don't provide dummy crash hooks for address_mapping functions o
          rename xfs_mmap to xfs_file_mmap
        o rewrite xfs_file_mmap in case of 2.4 to make use of
          xfs_file_vm_ops we can do this now since `filemap_nopage' is
          exported.
        o rewrite xfs_d_delete since in case of 2.4 it may not sleep, so
          when we are detecting that a xnode needs to be cleaned away
          return a 1 (that will remove the dentry directly in d_delete())
          also, use xfs_message_send() to notify useland that the node is
          unused, xfs_queue_inactive() have unwanted side-effects
        
        * xfs/linux/xfs/{xfs_fs.h,xfs_node.h}: change home cooked list to
        linux list macro, protect it will mutex

        * xfs/linux/xfs_dev.c: remove some ifdef's and use init_MUTEX
        directly
        * xfs/linux/xfs_common.c: if we have the flag GFP_FS, pass that in
        to malloc
        
        * xfs/linux/xfs/xfs_locl.h: if missing: define init_MUTEX, add
        function list_del_init() add list_for_each_safe,
        
        * configure.in: (linux): check for list_del_init
        * cf/linux-check-list-del-init.m4: new test

        * xfs/linux/Makefile.in: added modules_install: target
        * scripts/linux.gdb: xps
        
2002-01-25  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/mac/install/Arla: Don't start arla if there is no network
        (from William Uther <will+@cs.cmu.edu>)

        * arlad/arla-cli.c (arla_put): allow overwriting

2002-01-25  Love  <lha@stacken.kth.se>

        * configure.in: Disable (arlad/fbuf) mmap for darwin for now, we
        get 0 byte long files sometime.

2002-01-22  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * appl/lib/Makefile.in: Pass SYSCONFDIR

        * appl/lib/arlalib.c (arlalib_get_cred_krb): Added support for
          localauth
        o (key_to_key): Added

        * rxdef/common.h, rxdef/vldb.xg: Moved voltypes to common.h

2002-01-22  Love  <lha@stacken.kth.se>

        * configure.in: unbreak the --enable-smp for both linux and
        freebsd, for freebsd is plainly didn't work, and for linux it
        didn't autoprobe

        * lib/ko/kocell.c: (parse_simple_file): remove the newline from
        the row (if there is one)

        * rx/ChangeLog.c: enable softack and add dependences for rx_globs.o
        
        * rx/{rx.c,rx.h,rx_globs.h,rx_pkt.c,rx_pkt.h,rx_rdwr.c}:
        o Added softack support (acking packet on the recv side)
        o Make sure we send full-sized packets
        o Allocate packet buffer before we start, this way we will not use
          the two first fullsized packets
        o Added variables needed for congestion avoidance

        * rx/rx_clock.h: ident
        * rx/rx_user.c: do not use the mtu of the local loopback interface

2002-01-15  Love  <lha@stacken.kth.se>

        * lwp/plwp.h: waste memory and make lwp_debug an int (get away of
        signedness problem of char)
        * lwp/lwp_asm.h: waste memory and make lwp_debug an int (get away of
        signedness problem of char)
        
2002-01-11  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * rxdef/volumeserver.xg: Use string instead of char * in
        AFSVolSignalRestore and AFSVolSetIdsTypes

2002-01-07  Assar Westerlund  <assar@kth.se>

        * xfs/bsd/xfs/xfs_locl.h: FreeBSD 4.5 does define d_thread_t

        * xfs/bsd/xfs_message.c, xfs/bsd/xfs_node-bsd.c: handle struct
        mount.mnt_nvnodelist
        (FreeBSD 4.5 and higher)

        * configure.in (FreeBSD): never set smp flags to "no"
        (FreeBSD): test for struct mount.mnt_nvnodelist (the new name for
        mnt_vnodelist in 4.5 and above)

2002-01-05  Magnus Ahltorp  <ahltorp@nada.kth.se>

        * configure.in: Check for statvfs

For older ChangeLog entries see ChangeLog.2001