1 Changes with mod_fastcgi 2.2.5
3 *) seteuid() tweak for HP-UX 11. Milton L. Hankins [mlh@swl.msd.ray.com]
5 *) (Win32) More dynamic fixes.
7 *) (Win32) Eliminate the per application named pipe mutex. Its now
8 per process to keep the lib happy (the use _FCGI_MUTEX_ should
9 removed from the lib). This allows pipe-based applications to
10 now handle multiple simoultanous requests (one per process).
12 *) Increase the number of open FDs we look to close when spawning apps.
14 *) Prevent an assert from popping unnecessarily.
16 *) (Win32) Add support for interpretter scripts.
18 *) (Win32) Fix named pipe handling (problems with large responses).
20 *) (Win32) Remove the "can exec" check.
22 Changes with mod_fastcgi 2.2.4
24 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
25 Rob Saccoccio [robs@chelsea.net]
27 *) Remove request type restriction (GET and POST only) in order to support
28 Web DAV requests. Scott Robertson [sroberts@codeit.com]
30 *) Allow requests for URLs such as /server/some/other/qualifier to match a
31 FastCGI server defined as /server. This was done primarily for Java
32 Servlets, but is generically useful.
34 *) Change the comm between the PM and the request handlers from a regular
35 file to a pipe and an assortment of other dynamic fixes.
37 *) Log dynamic process termination scheduling and the resulting
38 process exit notification.
40 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
41 application process from being killed off due to low demand.
43 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
45 *) Fix bugs that prevented dynamic processes from being shutdown when
46 the load subsided. Lars Heete [heete@do.isst.fhg.de]
48 *) Prevent dynamic processes from being scheduled to be started before
49 the init-start-delay or restart-delay period expires. This prevents
50 the queuing of process start requests while an application which has
51 a long initialization period starts up.
53 *) When suexec is enabled, allow the process manager to become root again in
54 order to signal applications it spawned.
56 *) Change the "server started" log message level from INFO to WARN.
58 *) Fix dynamic server "Connection Refused" handling.
60 *) Fix demand-based dynamic process spawing (uses a more portable approach).
62 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
63 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
64 longer than this period. It applies to the initiation of connections as
65 well (and thus is similar to appConnTimeout). Default is 30 seconds.
67 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
68 is more platform portable/predicable.
70 *) Leave STDOUT and STDERR open to the main server error_log. This should
71 eliminate a pile of protocol errors and having to track down 3rd party
72 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
73 be fixed in the application to use the FastCGI I/O, it just allows these
74 applications to run when previously they'd crash and burn.
76 *) Add the pass-header arg to server directives.
78 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
79 is done for CGI) - at least until PR2580 is resolved.
81 *) Initialize the default (empty) environment in a more socially acceptable
84 *) Miscellaneous doc improvements based on notes I've collected up over the
87 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
88 that resulted in not or improperly terminated strings.
90 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
91 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
93 Changes with mod_fastcgi 2.2.2
95 *) Added support for blocking connect()s by setting appConnTimeout to 0.
96 Non-blocking connect()s (the default) can be troublesome on some platforms.
97 Its expected that blocking connect()s will become the default (and
98 non-blocking connect()s will become optional) in the next release.
100 *) Dump a compile time error if the version of Apache is too old.
102 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
103 to Apache releases prior to 1.3.6.
105 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
109 Always restart a failed dynamic application if it is the last instance.
110 This means there once started there will always be at least one process
111 instance of a dynamic application.
113 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
114 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
115 the PM to start (another) application instance may help empty it faster.
117 Change two sleep() calls to select() based snoozes because alarm() is in
118 effect and sleep() and alarm() don't always play nice together.
120 Fixed a couple of error messages.
122 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
123 was ignored and the default was always used.
125 *) Allow the -initial-env argument to be used to pass variables from the
126 Apache process environment to the FastCGI server (by specifying a
127 variable name without the "=" or a value). [suggested by
128 Martin Lichtin <lichtin@oanda.com>]
130 *) Cleanup some debug macros.
132 *) Improved script stderr handling. [based on suggestions by
133 David Birnbaum <davidb@chelsea.net>]
135 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
136 updates to the INSTALL doc.
138 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
139 5 to 100. Its still configurable with the -listen-queue-depth option.
140 This should help eliminate the FAQ - Why do I see "Connection Refused"
143 *) Fix a bug in FastCgiExternalServer that broke support for external servers
144 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
146 Changes with mod_fastcgi 2.2.1
148 *) Updates to the INSTALL doc to describe building as a DSO.
150 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
151 Apache was running ("apachectl restart" does this), the contents of the
152 dynamic directory (dynamic sockets and the mbox) would be blown away
155 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
156 Apache header file) to prevent a compile error on systems without
159 *) During auth requests, the subprocess_env table was left holding variables
160 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
161 were passed to other processing phases (such as CGI/FastCGI). The
162 subprocess_env is now restored to its pre-auth condition.
164 *) Added a FastCGI Authorizer Role compatibility mode which implements the
165 specification to the tee. Use -compat arg with any of the Auth server
166 directives. This is intended for new and existing FastCGI authorizer
167 applications that require compatibility with other server implementations.
169 *) Fix logging from the process manager. Previously, all calls to the
170 logging routines used NULL for the server_rec. This works fine in a dev
171 environment, but results in most of these messages being tossed by
172 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
174 *) Always setup fr->header so that in the event the FastCGI server
175 doesn't get a valid FastCGI protocol header over the wire, we
176 can still print our error message to the log from do_work().
178 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
179 defined (Don Locrasto [locrasto@iceminer.com]).
181 Changes with mod_fastcgi 2.2.0
183 *) Serious rewrite of mod_fastcgi.html. An example conf will have
184 to wait until another time - I'm burned out on docs updates.
186 *) Add to the mod_fastcgi to the server version string.
188 *) SCO doesn't like the const in the arg to inet_addr() and
191 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
193 *) Deleted redundant header files from fcgi.h.
195 *) Full path names are no longer required for directives which take
196 file names as arguments. Like other Apache directives you can
197 specify paths from server_root (don't start with a "/").
199 *) The fcgiKillMgr is gone. This intermediate process sat between
200 the Apache parent and the FastCGI Process Manager under Apache
201 1.2. The process manager tries to name itself fcgi_pm (it used
204 *) Authentication, Authorization, and Access phases are now
205 supported. I've bent the FastCgi spec a bit: as many of the standard
206 environment variables (that were easy) are sent (the spec says don't
207 send some by name), all headers (except Status) sent by the auth
208 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
209 as environment variables rather than just those prefixed by
210 "Variable-". Custom responses for auth failures aren't supported
211 (yet.. we'll see what the demand for this is like). In addition to
212 the FastCGI protocol defined environment variable "FCGI_ROLE" being
213 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
214 set indicating which of the three phases is being processed. See
215 the docs for info on the new per-directory directives:
217 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
218 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
219 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
221 *) The code has been broken into logical chunks making figuring it
222 out much easier. All of the #defines a user may want to change are
223 now in mod_fastcgi.h.
225 *) All the logging has been converted to the "supported" Apache
226 logging routines, ap_log_error() and ap_log_rerror(). Log entries
227 for request specific errors should now be directed to the correct
228 server errorlog (if your running more than one). Some attempt has
229 been made at setting appropriate log levels (comments of course are
230 welcome) and printing errno info when its of value. All of log
231 messages now have "FastCGI" in them (nice for grep).
233 *) All of the calls to Apache routines have been updated to the ap_
234 convention, eliminating the need for the compat header.
236 *) By default we no longer flush every piece of data we get from the
237 FastCGI server to the client. This allows the FastCGI server to
238 release without having to wait for the client flush to complete.
239 The old behaviour can revived with the -flush argument to AppClass.
241 *) I renamed some of the directives for consistency (the old names
244 AppClass -> FastCgiServer,
245 FCGIConfig -> FastCgiConfig,
246 ExternalAppClass -> FastCgiExternalServer
248 *) Directive arguments are no longer case sensitive.
250 *) The module now uses hard_timeout() rather than soft_timeout. This
251 means that the module will longjump out of the request if the
252 drop-dead timeout expires (set with Apache's Timeout directive, the
253 default is 5min) or if the client closes the connection (SIGPIPE).
254 This is more typical for a Apache modules and I'm not convinced we
255 properly handle all of the error cases well enough to use
256 soft_timeout (I can think of one place we don't anyway). This means
257 your FastCGI application can see SIGPIPE.
259 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
260 Apache. Without this the default behaviour is exit().
262 *) It should (I think) run now under DSO now.
264 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
266 *) The broken -initial-env bug ins 2.1b1 is fixed.
268 *) Maybe some other stuff.
272 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
273 commit notes and stick 'em in here.. when I get some free time. ;)
278 *** Originally from docs/README..
280 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
281 From: Stanley Gambarin <stanleyg@cs.bu.edu>
283 *) Yet more changes to the source distribution. Separated out
284 the Tcl dynamic string and buffer libraries into separate files.
285 Did the same for the OS library, however, it is not really an
286 abstraction... more like a bunch of wrappers. Added to DEVNOTES
287 a note about the problem of Apache not allowing including header
290 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
291 From: Stanley Gambarin <stanleyg@cs.bu.edu>
293 *) More changes to the installation scripts. Theoretically we
294 should be able to to use the same Makefile/installation script
295 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
296 layer to use the same source code for both 1.2.x and 1.3.x sources.
298 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
299 From: Stanley Gambarin <stanleyg@cs.bu.edu>
301 *) Created Makefile to compile the module into the archive library,
302 just like the proxy module. Also initial draft of the installation
303 script, which should work for both 1.2 and 1.3 sources of Apache.
305 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
306 From: Stanley Gambarin <stanleyg@cs.bu.edu>
308 *) More source reorg. Separated out some header files. Need much
309 more work on this. Updated the TODO file. Added fcgivers.h and
310 CHANGES file to track the history.
312 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
313 From: Stanley Gambarin <stanleyg@cs.bu.edu>
315 *) Source reorganization. This is done to provide a basis for later
316 functionality implementation. See TODO file for more information
317 of what possible future enhancements are possible/desired.
319 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
320 From: David MacKenzie <djm@va.pubnix.com>
322 *) Create the "dynamic" dir and "mbox" as the user specified in the
323 User and Group directives instead of as root. They are created with
324 restrictive permissions, and then the module checks to see whether the
325 user specified in the User and Group directives has read, write, and
326 execute permissions on them. Those permissions had been explicitly
327 denied by creating them as "root".
329 *) Some documentation was garbled or incomplete.
331 *) When a FastCGI application can't be execl'd, the code used a
332 value of errno that may have been stomped on by intervening system
333 calls. That happened on BSD/OS 2.1, where the error_log was
334 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
335 An intervening stdio call was setting errno as a side-effect, which
336 it has the right to do.
338 *) Erroneous fprintf arguments, a missing return value, an unused
339 function, missing declarations, and other problems detected by gcc
342 *) Duplicated code to create the lock file name merged into a single
343 function, closing several memory leaks.
345 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
346 FCGI app in a given dynamic class would live forever, even if it
347 hadn't been requested for weeks.
349 *) The test for whether to keep looking for dynamic app victims to
350 kill was backwards, so no dynamic apps were ever selected as
353 *) The killInterval and updateInterval were being ignored; the
354 sigsuspend() forced recalculations only when a child died,
355 instead of at the intervals specified.
357 *) When a dynamic app couldn't be started, the program SEGV'd when
358 trying to free the ipcAddrPtr twice.
360 *) Remove the dead lock file and socket when cleaning up after a
361 server whose last child has been killed, so the FastCGIHandler
362 isn't fooled into thinking there is still a process serving
365 *) Make file locking robust in the presence of signals.
367 *) Rename some badly named variables.
369 *) Fix typos in many comments.
371 *) Fix some memory and file descriptor leaks.
373 *) Make the blocking kill of a server closer to working.
375 *) More fixes to calculations in dynamic application management.
378 *** Originally from docs/README.OMI..
380 What's New: Version 2.0b1, 16 Apr 1997
382 *) Implemented a mechanism by which FastCGI applications are started by
383 the web server on the first request and continue running. Also, a
384 heuristics have been implemented to allow for dynamic killing of the
385 running FastCGI apps. The configuration options are supplied via
386 the new FCGIConfig directive, that is described in the
389 *) When performing internal redirect, since the original request's body
390 has already been read, do not allow the redirected request to think
391 that is has one [body].
393 *) More conditional compilation for OS/2.
395 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
397 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
399 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
400 force some output to be written in the period of inactivity. This
401 circumvents the Apache's buffering during the server-push.
403 *) Added a header-parser function placeholder for compatibility with an
406 *) Additional checks have been placed with regard to permissions of the
407 FastCGI processes. These include the existence of the file,
408 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
410 *) Implemented a restart cleanup, so that no parentlesses FastCGI
411 applications are left after Apache is restarted and/or terminated.
412 This was accomplished via two independent mechanisms, where the
413 first one prevents a start of the process manager on the first
414 reading of the configuration files. The second mechanism deals away
415 from Apache's implementation of fork() (via spawn_child()) and
416 implements its own forking, thus removing any dependencies on the
417 Apache to cleanup processes during termination phase.
419 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
420 itself in the file conf.h
422 What's New: Version 1.4.3, 15 Jan 1997
424 *) Fixed compilation warnings for various platforms, as well as
425 conditional compilation for OS/2.
427 What's New: Version 1.4.2, 12 Dec 1996
429 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
430 will proceed under this version of Apache web server.
432 *) As the result of porting, the "include virtual" construct of SSI
433 will now work correctly using either <Location> or <Directory>
436 What's New: Version 1.4.1, 4 Dec 1996
438 *) Checks have been removed from the ScanCGIHeaders that provided
439 for the presence of both Status and Location headers as being
440 an error. Contradictory to CGI/1.1 Internet Draft, both of
441 these headers are used by the current CGI applications.
443 What's New: Version 1.4, 22 Nov 1996
445 *) Added the -port option to AppClass, allowing TCP/IP communication.
446 Added the -socket option to AppClass, allowing Unix domain
447 communication via a configurable pathname.
449 *) Added the ExternalAppClass directive, allowing TCP/IP
450 communication with remote FastCGI applications.
452 *) The handler had its own code for generating HTTP response headers;
453 now it uses Apache's. This reduces the size of the module.
454 More importantly, it fixes the bug in which "include virtual"
455 sees the HTTP response headers.
457 *) The response header parser performed very little checking.
458 Now the parser enforces the guidelines in the CGI/1.1
459 Internet-Draft: Status and Location are mutually exclusive,
460 Location can only be a response to GET or HEAD,
461 CGI response headers can't be repeated, etc. (The CGI response
462 headers are the ones the handler interprets: Status, Location,
465 *) The response header parser used to miss CGI headers
466 with no whitespace after the colon, e.g. "Status:200 OK".
468 *) The response header parser sometimes interpreted the first line
469 of content as an RFC822 continuation line.
471 *) The handler implemented a nonstandard version of Location
472 which never used internal redirects. The handler
473 now attempts to implement Location as specified in the
474 CGI/1.1 Internet-Draft. The module documentation explains
477 *) Error log entries from the response header parser were
478 pretty uninformative; they are better now. If a header is
479 malformed, the log entry includes it. If the headers are
480 unterminated rather than malformed, the log entry says that,
481 and says how many bytes were received from the app.
483 *) When the application manager forked a new process, and that
484 process ran into trouble before executing the first instruction
485 of application code, the process used to exit with status =
486 errno. This made certain configuration problems (e.g. incorrect
487 file permissions when server parent is root) quite difficult
488 to pin down. Now the failing child process opens up the error log
489 and writes an informative entry before exiting.
491 *) The module now correctly handles a slash at the end of the
492 DocumentRoot directive. This was a one-line fix.
494 What's New: Version 1.3.3, 17 Oct 1996
496 *) The module now registers its request handler under the name
497 fastcgi-script in addition to the name application/x-httpd-fcgi.
498 This was a one-line addition to the module, but had quite
499 a large impact on the documentation and sample configuration.
501 What's New: Version 1.3.2, 27 Sept 1996
503 *) On some systems (SunOS, Linux?), fopen for append has a bug
504 that strikes when two processes append to the same file.
505 This bug causes the process manager to corrupt the error log.
506 Work around the bug by calling open, then fdopen. A patch
507 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
508 (Reported by Bob Ramstad.)
510 What's New: Version 1.3.1, 17 Sept 1996
512 *) On Linux and SunOS, the Apache default user_id (-1) is not a
513 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
514 Scott Langley, others.)
516 *) On some systems (Linux, some Solaris?, Irix?), connect requires
517 write access to a Unix Domain socket, so provide it. (Reported
518 by Scott Langley, freeform@wired.com, others.)
520 *) If you hit a .fcg file, but there's no AppClass defined, the
521 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
524 What's New: Version 1.3, 4 Sept 1996
526 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
528 *) Module violated the Apache buff abstraction.
529 Fixed. Might now work with SSL (not tested).
530 As part of the fix, eliminated support for
531 Apache 1.0x versions.
533 *) Module failed to chmod the Unix domain listening sockets
534 it created, so protections were set according to the current
537 *) Module forked too many process manager processes: One
538 per AppClass. When the Apache parent ran as root, these
539 process manager processes ran as root. New process
540 manager is a single process (only started if AppClass is
541 used) and runs with same privileges as other children.
542 The new process manager doesn't do any polling, so
543 there's less system overhead than before.
545 *) AppClass insisted on getting at least two arguments. Fixed.
547 *) Module used setjmp/longjmp, causing compiler warnings
548 on some platforms (e.g. Linux.) The module no longer
551 *) Module created listening sockets in /tmp, where they were
552 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
553 directive to give control over the location of listening
556 *) Module wrote error log entries without a timestamp. Fixed.
558 *) AppClass directive wrote error messages to stderr in addition
559 to returning a char * message to the Apache core. Fixed.
561 What's New: Version 1.2, 3 June 1996
563 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
566 Add version string: APACHEVERSION macro to mod_fastcgi.c
568 *) chown FastCGI socket to user_id and setuid to user_id for app
571 *) Modify GetFromStream() by having it call an OS dependent
572 function GetStreamSize(FILE *) which uses FILE internal data member.
573 Linux users might need to modify GetStreamSize().
575 What's New: Version 1.1, 10 May 1996
577 *) If you specify a non-existent executable in the AppClass directive,
578 or if the file exists and it does not have execute permission you
579 get a constant stream of error messages telling you that "program
580 terminated due to a signal".
582 *) The mod_fastcgi module should use the standard Apache error logging
583 facility instead of writing to stderr.
585 Version 1.0, 30 April 1996