3 *) Docs cleanup. Andrew Benham [adsb@bigfoot.com]
5 *) Added code so if the last instance of a dynamic application died without
6 provocation, then don't restart it if singleThreshold > 0 (i.e. if the
7 configuration allows the last instance to be killed, then allow it to die).
8 Andrew Benham [adsb@bigfoot.com]
10 *) Fix the loadFactor calculation used to determine when dyanmic
11 applications could be killed off due to low demand [adsb@bigfoot.com].
13 *) Fix a deadlock condition that could occur with Win32 named
16 *) Fix a potential deadlock condition when FastCGI application
17 sent responses while still reading the client request (POST data).
21 *) Allow absolute pathnames in the -socket argument. Suggested by
22 Christian Jaeger [christian.jaeger@sl.ethz.ch].
24 *) Don't invoke suexec when the user/group for the fastcgi application
25 is the same as the apache main server. This is consistent with
26 apache's suexec handling. Suggested by Nikolaus Rath [Nikolaus@rath.org].
28 *) Reset the apache drop dead timer upon successful read or writes
29 to/from the client. This eliminates timeouts that were occuring
30 during the large file transfers to/from slow clients.
32 *) Support generic wrappers such as cgiwrap by eliminating dependencies
33 on Apache's SUEXEC, renaming the FastCgiSuexec directive
34 FastCgiWrapper and eliminating any checks regarding the target
35 application (this is the repsonibility of the wrapper).
37 *) Fix a nasty bug that occurred when a client aborted a POST request
38 before the connection to a dynamic FastCGI application was opened.
39 The application's lock file descriptor wasn't setup but was being
40 closed which resulted in FD0 being closed. Normally this is open to
41 /dev/null and should pose no problems except that because the FD
42 was available it was being returned by Apache's accept(). This
43 caused it to be registered for pool cleanup. mod_cgi though moves
44 the CGI stdin pipe to FD0 and thus it was getting waxed during pool
45 cleanup. Problem identified by checksum@163.net.
47 Changes with mod_fastcgi 2.2.8
49 *) Eliminate the concept of disabled applications. If a failure
50 occurs trying to setup an application (e.g. bind() error) its
51 tried repeatedly every init-start-delay seconds.
53 *) Tweak to Makefile.tmpl to support DSOs. Dave Hill [ddhill@zk3.dec.com]
55 Changes with mod_fastcgi 2.2.6
57 *) Shutdown the PM when Apache appears to have disappeared.
59 *) seteuid() tweak for HP-UX 11. Milton L. Hankins [mlh@swl.msd.ray.com]
61 *) (Win32) More dynamic fixes.
63 *) (Win32) Eliminate the per application named pipe mutex. Its now
64 per process to keep the lib happy (the use _FCGI_MUTEX_ should
65 removed from the lib). This allows pipe-based applications to
66 now handle multiple simoultanous requests (one per process).
68 *) Increase the number of open FDs we look to close when spawning apps.
70 *) Prevent an assert from popping unnecessarily.
72 *) (Win32) Add support for interpretter scripts.
74 *) (Win32) Fix named pipe handling (problems with large responses).
76 *) (Win32) Remove the "can exec" check.
78 Changes with mod_fastcgi 2.2.4
80 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
81 Rob Saccoccio [robs@chelsea.net]
83 *) Remove request type restriction (GET and POST only) in order to support
84 Web DAV requests. Scott Robertson [sroberts@codeit.com]
86 *) Allow requests for URLs such as /server/some/other/qualifier to match a
87 FastCGI server defined as /server. This was done primarily for Java
88 Servlets, but is generically useful.
90 *) Change the comm between the PM and the request handlers from a regular
91 file to a pipe and an assortment of other dynamic fixes.
93 *) Log dynamic process termination scheduling and the resulting
94 process exit notification.
96 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
97 application process from being killed off due to low demand.
99 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
101 *) Fix bugs that prevented dynamic processes from being shutdown when
102 the load subsided. Lars Heete [heete@do.isst.fhg.de]
104 *) Prevent dynamic processes from being scheduled to be started before
105 the init-start-delay or restart-delay period expires. This prevents
106 the queuing of process start requests while an application which has
107 a long initialization period starts up.
109 *) When suexec is enabled, allow the process manager to become root again in
110 order to signal applications it spawned.
112 *) Change the "server started" log message level from INFO to WARN.
114 *) Fix dynamic server "Connection Refused" handling.
116 *) Fix demand-based dynamic process spawing (uses a more portable approach).
118 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
119 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
120 longer than this period. It applies to the initiation of connections as
121 well (and thus is similar to appConnTimeout). Default is 30 seconds.
123 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
124 is more platform portable/predicable.
126 *) Leave STDOUT and STDERR open to the main server error_log. This should
127 eliminate a pile of protocol errors and having to track down 3rd party
128 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
129 be fixed in the application to use the FastCGI I/O, it just allows these
130 applications to run when previously they'd crash and burn.
132 *) Add the pass-header arg to server directives.
134 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
135 is done for CGI) - at least until PR2580 is resolved.
137 *) Initialize the default (empty) environment in a more socially acceptable
140 *) Miscellaneous doc improvements based on notes I've collected up over the
143 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
144 that resulted in not or improperly terminated strings.
146 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
147 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
149 Changes with mod_fastcgi 2.2.2
151 *) Added support for blocking connect()s by setting appConnTimeout to 0.
152 Non-blocking connect()s (the default) can be troublesome on some platforms.
153 Its expected that blocking connect()s will become the default (and
154 non-blocking connect()s will become optional) in the next release.
156 *) Dump a compile time error if the version of Apache is too old.
158 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
159 to Apache releases prior to 1.3.6.
161 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
165 Always restart a failed dynamic application if it is the last instance.
166 This means there once started there will always be at least one process
167 instance of a dynamic application.
169 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
170 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
171 the PM to start (another) application instance may help empty it faster.
173 Change two sleep() calls to select() based snoozes because alarm() is in
174 effect and sleep() and alarm() don't always play nice together.
176 Fixed a couple of error messages.
178 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
179 was ignored and the default was always used.
181 *) Allow the -initial-env argument to be used to pass variables from the
182 Apache process environment to the FastCGI server (by specifying a
183 variable name without the "=" or a value). [suggested by
184 Martin Lichtin <lichtin@oanda.com>]
186 *) Cleanup some debug macros.
188 *) Improved script stderr handling. [based on suggestions by
189 David Birnbaum <davidb@chelsea.net>]
191 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
192 updates to the INSTALL doc.
194 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
195 5 to 100. Its still configurable with the -listen-queue-depth option.
196 This should help eliminate the FAQ - Why do I see "Connection Refused"
199 *) Fix a bug in FastCgiExternalServer that broke support for external servers
200 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
202 Changes with mod_fastcgi 2.2.1
204 *) Updates to the INSTALL doc to describe building as a DSO.
206 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
207 Apache was running ("apachectl restart" does this), the contents of the
208 dynamic directory (dynamic sockets and the mbox) would be blown away
211 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
212 Apache header file) to prevent a compile error on systems without
215 *) During auth requests, the subprocess_env table was left holding variables
216 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
217 were passed to other processing phases (such as CGI/FastCGI). The
218 subprocess_env is now restored to its pre-auth condition.
220 *) Added a FastCGI Authorizer Role compatibility mode which implements the
221 specification to the tee. Use -compat arg with any of the Auth server
222 directives. This is intended for new and existing FastCGI authorizer
223 applications that require compatibility with other server implementations.
225 *) Fix logging from the process manager. Previously, all calls to the
226 logging routines used NULL for the server_rec. This works fine in a dev
227 environment, but results in most of these messages being tossed by
228 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
230 *) Always setup fr->header so that in the event the FastCGI server
231 doesn't get a valid FastCGI protocol header over the wire, we
232 can still print our error message to the log from do_work().
234 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
235 defined (Don Locrasto [locrasto@iceminer.com]).
237 Changes with mod_fastcgi 2.2.0
239 *) Serious rewrite of mod_fastcgi.html. An example conf will have
240 to wait until another time - I'm burned out on docs updates.
242 *) Add to the mod_fastcgi to the server version string.
244 *) SCO doesn't like the const in the arg to inet_addr() and
247 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
249 *) Deleted redundant header files from fcgi.h.
251 *) Full path names are no longer required for directives which take
252 file names as arguments. Like other Apache directives you can
253 specify paths from server_root (don't start with a "/").
255 *) The fcgiKillMgr is gone. This intermediate process sat between
256 the Apache parent and the FastCGI Process Manager under Apache
257 1.2. The process manager tries to name itself fcgi_pm (it used
260 *) Authentication, Authorization, and Access phases are now
261 supported. I've bent the FastCgi spec a bit: as many of the standard
262 environment variables (that were easy) are sent (the spec says don't
263 send some by name), all headers (except Status) sent by the auth
264 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
265 as environment variables rather than just those prefixed by
266 "Variable-". Custom responses for auth failures aren't supported
267 (yet.. we'll see what the demand for this is like). In addition to
268 the FastCGI protocol defined environment variable "FCGI_ROLE" being
269 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
270 set indicating which of the three phases is being processed. See
271 the docs for info on the new per-directory directives:
273 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
274 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
275 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
277 *) The code has been broken into logical chunks making figuring it
278 out much easier. All of the #defines a user may want to change are
279 now in mod_fastcgi.h.
281 *) All the logging has been converted to the "supported" Apache
282 logging routines, ap_log_error() and ap_log_rerror(). Log entries
283 for request specific errors should now be directed to the correct
284 server errorlog (if your running more than one). Some attempt has
285 been made at setting appropriate log levels (comments of course are
286 welcome) and printing errno info when its of value. All of log
287 messages now have "FastCGI" in them (nice for grep).
289 *) All of the calls to Apache routines have been updated to the ap_
290 convention, eliminating the need for the compat header.
292 *) By default we no longer flush every piece of data we get from the
293 FastCGI server to the client. This allows the FastCGI server to
294 release without having to wait for the client flush to complete.
295 The old behaviour can revived with the -flush argument to AppClass.
297 *) I renamed some of the directives for consistency (the old names
300 AppClass -> FastCgiServer,
301 FCGIConfig -> FastCgiConfig,
302 ExternalAppClass -> FastCgiExternalServer
304 *) Directive arguments are no longer case sensitive.
306 *) The module now uses hard_timeout() rather than soft_timeout. This
307 means that the module will longjump out of the request if the
308 drop-dead timeout expires (set with Apache's Timeout directive, the
309 default is 5min) or if the client closes the connection (SIGPIPE).
310 This is more typical for a Apache modules and I'm not convinced we
311 properly handle all of the error cases well enough to use
312 soft_timeout (I can think of one place we don't anyway). This means
313 your FastCGI application can see SIGPIPE.
315 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
316 Apache. Without this the default behaviour is exit().
318 *) It should (I think) run now under DSO now.
320 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
322 *) The broken -initial-env bug ins 2.1b1 is fixed.
324 *) Maybe some other stuff.
328 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
329 commit notes and stick 'em in here.. when I get some free time. ;)
334 *** Originally from docs/README..
336 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
337 From: Stanley Gambarin <stanleyg@cs.bu.edu>
339 *) Yet more changes to the source distribution. Separated out
340 the Tcl dynamic string and buffer libraries into separate files.
341 Did the same for the OS library, however, it is not really an
342 abstraction... more like a bunch of wrappers. Added to DEVNOTES
343 a note about the problem of Apache not allowing including header
346 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
347 From: Stanley Gambarin <stanleyg@cs.bu.edu>
349 *) More changes to the installation scripts. Theoretically we
350 should be able to to use the same Makefile/installation script
351 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
352 layer to use the same source code for both 1.2.x and 1.3.x sources.
354 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
355 From: Stanley Gambarin <stanleyg@cs.bu.edu>
357 *) Created Makefile to compile the module into the archive library,
358 just like the proxy module. Also initial draft of the installation
359 script, which should work for both 1.2 and 1.3 sources of Apache.
361 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
362 From: Stanley Gambarin <stanleyg@cs.bu.edu>
364 *) More source reorg. Separated out some header files. Need much
365 more work on this. Updated the TODO file. Added fcgivers.h and
366 CHANGES file to track the history.
368 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
369 From: Stanley Gambarin <stanleyg@cs.bu.edu>
371 *) Source reorganization. This is done to provide a basis for later
372 functionality implementation. See TODO file for more information
373 of what possible future enhancements are possible/desired.
375 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
376 From: David MacKenzie <djm@va.pubnix.com>
378 *) Create the "dynamic" dir and "mbox" as the user specified in the
379 User and Group directives instead of as root. They are created with
380 restrictive permissions, and then the module checks to see whether the
381 user specified in the User and Group directives has read, write, and
382 execute permissions on them. Those permissions had been explicitly
383 denied by creating them as "root".
385 *) Some documentation was garbled or incomplete.
387 *) When a FastCGI application can't be execl'd, the code used a
388 value of errno that may have been stomped on by intervening system
389 calls. That happened on BSD/OS 2.1, where the error_log was
390 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
391 An intervening stdio call was setting errno as a side-effect, which
392 it has the right to do.
394 *) Erroneous fprintf arguments, a missing return value, an unused
395 function, missing declarations, and other problems detected by gcc
398 *) Duplicated code to create the lock file name merged into a single
399 function, closing several memory leaks.
401 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
402 FCGI app in a given dynamic class would live forever, even if it
403 hadn't been requested for weeks.
405 *) The test for whether to keep looking for dynamic app victims to
406 kill was backwards, so no dynamic apps were ever selected as
409 *) The killInterval and updateInterval were being ignored; the
410 sigsuspend() forced recalculations only when a child died,
411 instead of at the intervals specified.
413 *) When a dynamic app couldn't be started, the program SEGV'd when
414 trying to free the ipcAddrPtr twice.
416 *) Remove the dead lock file and socket when cleaning up after a
417 server whose last child has been killed, so the FastCGIHandler
418 isn't fooled into thinking there is still a process serving
421 *) Make file locking robust in the presence of signals.
423 *) Rename some badly named variables.
425 *) Fix typos in many comments.
427 *) Fix some memory and file descriptor leaks.
429 *) Make the blocking kill of a server closer to working.
431 *) More fixes to calculations in dynamic application management.
434 *** Originally from docs/README.OMI..
436 What's New: Version 2.0b1, 16 Apr 1997
438 *) Implemented a mechanism by which FastCGI applications are started by
439 the web server on the first request and continue running. Also, a
440 heuristics have been implemented to allow for dynamic killing of the
441 running FastCGI apps. The configuration options are supplied via
442 the new FCGIConfig directive, that is described in the
445 *) When performing internal redirect, since the original request's body
446 has already been read, do not allow the redirected request to think
447 that is has one [body].
449 *) More conditional compilation for OS/2.
451 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
453 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
455 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
456 force some output to be written in the period of inactivity. This
457 circumvents the Apache's buffering during the server-push.
459 *) Added a header-parser function placeholder for compatibility with an
462 *) Additional checks have been placed with regard to permissions of the
463 FastCGI processes. These include the existence of the file,
464 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
466 *) Implemented a restart cleanup, so that no parentlesses FastCGI
467 applications are left after Apache is restarted and/or terminated.
468 This was accomplished via two independent mechanisms, where the
469 first one prevents a start of the process manager on the first
470 reading of the configuration files. The second mechanism deals away
471 from Apache's implementation of fork() (via spawn_child()) and
472 implements its own forking, thus removing any dependencies on the
473 Apache to cleanup processes during termination phase.
475 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
476 itself in the file conf.h
478 What's New: Version 1.4.3, 15 Jan 1997
480 *) Fixed compilation warnings for various platforms, as well as
481 conditional compilation for OS/2.
483 What's New: Version 1.4.2, 12 Dec 1996
485 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
486 will proceed under this version of Apache web server.
488 *) As the result of porting, the "include virtual" construct of SSI
489 will now work correctly using either <Location> or <Directory>
492 What's New: Version 1.4.1, 4 Dec 1996
494 *) Checks have been removed from the ScanCGIHeaders that provided
495 for the presence of both Status and Location headers as being
496 an error. Contradictory to CGI/1.1 Internet Draft, both of
497 these headers are used by the current CGI applications.
499 What's New: Version 1.4, 22 Nov 1996
501 *) Added the -port option to AppClass, allowing TCP/IP communication.
502 Added the -socket option to AppClass, allowing Unix domain
503 communication via a configurable pathname.
505 *) Added the ExternalAppClass directive, allowing TCP/IP
506 communication with remote FastCGI applications.
508 *) The handler had its own code for generating HTTP response headers;
509 now it uses Apache's. This reduces the size of the module.
510 More importantly, it fixes the bug in which "include virtual"
511 sees the HTTP response headers.
513 *) The response header parser performed very little checking.
514 Now the parser enforces the guidelines in the CGI/1.1
515 Internet-Draft: Status and Location are mutually exclusive,
516 Location can only be a response to GET or HEAD,
517 CGI response headers can't be repeated, etc. (The CGI response
518 headers are the ones the handler interprets: Status, Location,
521 *) The response header parser used to miss CGI headers
522 with no whitespace after the colon, e.g. "Status:200 OK".
524 *) The response header parser sometimes interpreted the first line
525 of content as an RFC822 continuation line.
527 *) The handler implemented a nonstandard version of Location
528 which never used internal redirects. The handler
529 now attempts to implement Location as specified in the
530 CGI/1.1 Internet-Draft. The module documentation explains
533 *) Error log entries from the response header parser were
534 pretty uninformative; they are better now. If a header is
535 malformed, the log entry includes it. If the headers are
536 unterminated rather than malformed, the log entry says that,
537 and says how many bytes were received from the app.
539 *) When the application manager forked a new process, and that
540 process ran into trouble before executing the first instruction
541 of application code, the process used to exit with status =
542 errno. This made certain configuration problems (e.g. incorrect
543 file permissions when server parent is root) quite difficult
544 to pin down. Now the failing child process opens up the error log
545 and writes an informative entry before exiting.
547 *) The module now correctly handles a slash at the end of the
548 DocumentRoot directive. This was a one-line fix.
550 What's New: Version 1.3.3, 17 Oct 1996
552 *) The module now registers its request handler under the name
553 fastcgi-script in addition to the name application/x-httpd-fcgi.
554 This was a one-line addition to the module, but had quite
555 a large impact on the documentation and sample configuration.
557 What's New: Version 1.3.2, 27 Sept 1996
559 *) On some systems (SunOS, Linux?), fopen for append has a bug
560 that strikes when two processes append to the same file.
561 This bug causes the process manager to corrupt the error log.
562 Work around the bug by calling open, then fdopen. A patch
563 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
564 (Reported by Bob Ramstad.)
566 What's New: Version 1.3.1, 17 Sept 1996
568 *) On Linux and SunOS, the Apache default user_id (-1) is not a
569 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
570 Scott Langley, others.)
572 *) On some systems (Linux, some Solaris?, Irix?), connect requires
573 write access to a Unix Domain socket, so provide it. (Reported
574 by Scott Langley, freeform@wired.com, others.)
576 *) If you hit a .fcg file, but there's no AppClass defined, the
577 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
580 What's New: Version 1.3, 4 Sept 1996
582 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
584 *) Module violated the Apache buff abstraction.
585 Fixed. Might now work with SSL (not tested).
586 As part of the fix, eliminated support for
587 Apache 1.0x versions.
589 *) Module failed to chmod the Unix domain listening sockets
590 it created, so protections were set according to the current
593 *) Module forked too many process manager processes: One
594 per AppClass. When the Apache parent ran as root, these
595 process manager processes ran as root. New process
596 manager is a single process (only started if AppClass is
597 used) and runs with same privileges as other children.
598 The new process manager doesn't do any polling, so
599 there's less system overhead than before.
601 *) AppClass insisted on getting at least two arguments. Fixed.
603 *) Module used setjmp/longjmp, causing compiler warnings
604 on some platforms (e.g. Linux.) The module no longer
607 *) Module created listening sockets in /tmp, where they were
608 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
609 directive to give control over the location of listening
612 *) Module wrote error log entries without a timestamp. Fixed.
614 *) AppClass directive wrote error messages to stderr in addition
615 to returning a char * message to the Apache core. Fixed.
617 What's New: Version 1.2, 3 June 1996
619 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
622 Add version string: APACHEVERSION macro to mod_fastcgi.c
624 *) chown FastCGI socket to user_id and setuid to user_id for app
627 *) Modify GetFromStream() by having it call an OS dependent
628 function GetStreamSize(FILE *) which uses FILE internal data member.
629 Linux users might need to modify GetStreamSize().
631 What's New: Version 1.1, 10 May 1996
633 *) If you specify a non-existent executable in the AppClass directive,
634 or if the file exists and it does not have execute permission you
635 get a constant stream of error messages telling you that "program
636 terminated due to a signal".
638 *) The mod_fastcgi module should use the standard Apache error logging
639 facility instead of writing to stderr.
641 Version 1.0, 30 April 1996