1 Changes with mod_fastcgi 2.2.7
3 *) Tweak to Makefile.tmpl to support DSOs. Dave Hill [ddhill@zk3.dec.com]
5 Changes with mod_fastcgi 2.2.6
7 *) Shutdown the PM when Apache appears to have disappeared.
9 *) seteuid() tweak for HP-UX 11. Milton L. Hankins [mlh@swl.msd.ray.com]
11 *) (Win32) More dynamic fixes.
13 *) (Win32) Eliminate the per application named pipe mutex. Its now
14 per process to keep the lib happy (the use _FCGI_MUTEX_ should
15 removed from the lib). This allows pipe-based applications to
16 now handle multiple simoultanous requests (one per process).
18 *) Increase the number of open FDs we look to close when spawning apps.
20 *) Prevent an assert from popping unnecessarily.
22 *) (Win32) Add support for interpretter scripts.
24 *) (Win32) Fix named pipe handling (problems with large responses).
26 *) (Win32) Remove the "can exec" check.
28 Changes with mod_fastcgi 2.2.4
30 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
31 Rob Saccoccio [robs@chelsea.net]
33 *) Remove request type restriction (GET and POST only) in order to support
34 Web DAV requests. Scott Robertson [sroberts@codeit.com]
36 *) Allow requests for URLs such as /server/some/other/qualifier to match a
37 FastCGI server defined as /server. This was done primarily for Java
38 Servlets, but is generically useful.
40 *) Change the comm between the PM and the request handlers from a regular
41 file to a pipe and an assortment of other dynamic fixes.
43 *) Log dynamic process termination scheduling and the resulting
44 process exit notification.
46 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
47 application process from being killed off due to low demand.
49 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
51 *) Fix bugs that prevented dynamic processes from being shutdown when
52 the load subsided. Lars Heete [heete@do.isst.fhg.de]
54 *) Prevent dynamic processes from being scheduled to be started before
55 the init-start-delay or restart-delay period expires. This prevents
56 the queuing of process start requests while an application which has
57 a long initialization period starts up.
59 *) When suexec is enabled, allow the process manager to become root again in
60 order to signal applications it spawned.
62 *) Change the "server started" log message level from INFO to WARN.
64 *) Fix dynamic server "Connection Refused" handling.
66 *) Fix demand-based dynamic process spawing (uses a more portable approach).
68 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
69 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
70 longer than this period. It applies to the initiation of connections as
71 well (and thus is similar to appConnTimeout). Default is 30 seconds.
73 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
74 is more platform portable/predicable.
76 *) Leave STDOUT and STDERR open to the main server error_log. This should
77 eliminate a pile of protocol errors and having to track down 3rd party
78 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
79 be fixed in the application to use the FastCGI I/O, it just allows these
80 applications to run when previously they'd crash and burn.
82 *) Add the pass-header arg to server directives.
84 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
85 is done for CGI) - at least until PR2580 is resolved.
87 *) Initialize the default (empty) environment in a more socially acceptable
90 *) Miscellaneous doc improvements based on notes I've collected up over the
93 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
94 that resulted in not or improperly terminated strings.
96 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
97 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
99 Changes with mod_fastcgi 2.2.2
101 *) Added support for blocking connect()s by setting appConnTimeout to 0.
102 Non-blocking connect()s (the default) can be troublesome on some platforms.
103 Its expected that blocking connect()s will become the default (and
104 non-blocking connect()s will become optional) in the next release.
106 *) Dump a compile time error if the version of Apache is too old.
108 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
109 to Apache releases prior to 1.3.6.
111 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
115 Always restart a failed dynamic application if it is the last instance.
116 This means there once started there will always be at least one process
117 instance of a dynamic application.
119 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
120 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
121 the PM to start (another) application instance may help empty it faster.
123 Change two sleep() calls to select() based snoozes because alarm() is in
124 effect and sleep() and alarm() don't always play nice together.
126 Fixed a couple of error messages.
128 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
129 was ignored and the default was always used.
131 *) Allow the -initial-env argument to be used to pass variables from the
132 Apache process environment to the FastCGI server (by specifying a
133 variable name without the "=" or a value). [suggested by
134 Martin Lichtin <lichtin@oanda.com>]
136 *) Cleanup some debug macros.
138 *) Improved script stderr handling. [based on suggestions by
139 David Birnbaum <davidb@chelsea.net>]
141 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
142 updates to the INSTALL doc.
144 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
145 5 to 100. Its still configurable with the -listen-queue-depth option.
146 This should help eliminate the FAQ - Why do I see "Connection Refused"
149 *) Fix a bug in FastCgiExternalServer that broke support for external servers
150 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
152 Changes with mod_fastcgi 2.2.1
154 *) Updates to the INSTALL doc to describe building as a DSO.
156 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
157 Apache was running ("apachectl restart" does this), the contents of the
158 dynamic directory (dynamic sockets and the mbox) would be blown away
161 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
162 Apache header file) to prevent a compile error on systems without
165 *) During auth requests, the subprocess_env table was left holding variables
166 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
167 were passed to other processing phases (such as CGI/FastCGI). The
168 subprocess_env is now restored to its pre-auth condition.
170 *) Added a FastCGI Authorizer Role compatibility mode which implements the
171 specification to the tee. Use -compat arg with any of the Auth server
172 directives. This is intended for new and existing FastCGI authorizer
173 applications that require compatibility with other server implementations.
175 *) Fix logging from the process manager. Previously, all calls to the
176 logging routines used NULL for the server_rec. This works fine in a dev
177 environment, but results in most of these messages being tossed by
178 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
180 *) Always setup fr->header so that in the event the FastCGI server
181 doesn't get a valid FastCGI protocol header over the wire, we
182 can still print our error message to the log from do_work().
184 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
185 defined (Don Locrasto [locrasto@iceminer.com]).
187 Changes with mod_fastcgi 2.2.0
189 *) Serious rewrite of mod_fastcgi.html. An example conf will have
190 to wait until another time - I'm burned out on docs updates.
192 *) Add to the mod_fastcgi to the server version string.
194 *) SCO doesn't like the const in the arg to inet_addr() and
197 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
199 *) Deleted redundant header files from fcgi.h.
201 *) Full path names are no longer required for directives which take
202 file names as arguments. Like other Apache directives you can
203 specify paths from server_root (don't start with a "/").
205 *) The fcgiKillMgr is gone. This intermediate process sat between
206 the Apache parent and the FastCGI Process Manager under Apache
207 1.2. The process manager tries to name itself fcgi_pm (it used
210 *) Authentication, Authorization, and Access phases are now
211 supported. I've bent the FastCgi spec a bit: as many of the standard
212 environment variables (that were easy) are sent (the spec says don't
213 send some by name), all headers (except Status) sent by the auth
214 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
215 as environment variables rather than just those prefixed by
216 "Variable-". Custom responses for auth failures aren't supported
217 (yet.. we'll see what the demand for this is like). In addition to
218 the FastCGI protocol defined environment variable "FCGI_ROLE" being
219 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
220 set indicating which of the three phases is being processed. See
221 the docs for info on the new per-directory directives:
223 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
224 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
225 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
227 *) The code has been broken into logical chunks making figuring it
228 out much easier. All of the #defines a user may want to change are
229 now in mod_fastcgi.h.
231 *) All the logging has been converted to the "supported" Apache
232 logging routines, ap_log_error() and ap_log_rerror(). Log entries
233 for request specific errors should now be directed to the correct
234 server errorlog (if your running more than one). Some attempt has
235 been made at setting appropriate log levels (comments of course are
236 welcome) and printing errno info when its of value. All of log
237 messages now have "FastCGI" in them (nice for grep).
239 *) All of the calls to Apache routines have been updated to the ap_
240 convention, eliminating the need for the compat header.
242 *) By default we no longer flush every piece of data we get from the
243 FastCGI server to the client. This allows the FastCGI server to
244 release without having to wait for the client flush to complete.
245 The old behaviour can revived with the -flush argument to AppClass.
247 *) I renamed some of the directives for consistency (the old names
250 AppClass -> FastCgiServer,
251 FCGIConfig -> FastCgiConfig,
252 ExternalAppClass -> FastCgiExternalServer
254 *) Directive arguments are no longer case sensitive.
256 *) The module now uses hard_timeout() rather than soft_timeout. This
257 means that the module will longjump out of the request if the
258 drop-dead timeout expires (set with Apache's Timeout directive, the
259 default is 5min) or if the client closes the connection (SIGPIPE).
260 This is more typical for a Apache modules and I'm not convinced we
261 properly handle all of the error cases well enough to use
262 soft_timeout (I can think of one place we don't anyway). This means
263 your FastCGI application can see SIGPIPE.
265 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
266 Apache. Without this the default behaviour is exit().
268 *) It should (I think) run now under DSO now.
270 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
272 *) The broken -initial-env bug ins 2.1b1 is fixed.
274 *) Maybe some other stuff.
278 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
279 commit notes and stick 'em in here.. when I get some free time. ;)
284 *** Originally from docs/README..
286 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
287 From: Stanley Gambarin <stanleyg@cs.bu.edu>
289 *) Yet more changes to the source distribution. Separated out
290 the Tcl dynamic string and buffer libraries into separate files.
291 Did the same for the OS library, however, it is not really an
292 abstraction... more like a bunch of wrappers. Added to DEVNOTES
293 a note about the problem of Apache not allowing including header
296 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
297 From: Stanley Gambarin <stanleyg@cs.bu.edu>
299 *) More changes to the installation scripts. Theoretically we
300 should be able to to use the same Makefile/installation script
301 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
302 layer to use the same source code for both 1.2.x and 1.3.x sources.
304 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
305 From: Stanley Gambarin <stanleyg@cs.bu.edu>
307 *) Created Makefile to compile the module into the archive library,
308 just like the proxy module. Also initial draft of the installation
309 script, which should work for both 1.2 and 1.3 sources of Apache.
311 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
312 From: Stanley Gambarin <stanleyg@cs.bu.edu>
314 *) More source reorg. Separated out some header files. Need much
315 more work on this. Updated the TODO file. Added fcgivers.h and
316 CHANGES file to track the history.
318 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
319 From: Stanley Gambarin <stanleyg@cs.bu.edu>
321 *) Source reorganization. This is done to provide a basis for later
322 functionality implementation. See TODO file for more information
323 of what possible future enhancements are possible/desired.
325 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
326 From: David MacKenzie <djm@va.pubnix.com>
328 *) Create the "dynamic" dir and "mbox" as the user specified in the
329 User and Group directives instead of as root. They are created with
330 restrictive permissions, and then the module checks to see whether the
331 user specified in the User and Group directives has read, write, and
332 execute permissions on them. Those permissions had been explicitly
333 denied by creating them as "root".
335 *) Some documentation was garbled or incomplete.
337 *) When a FastCGI application can't be execl'd, the code used a
338 value of errno that may have been stomped on by intervening system
339 calls. That happened on BSD/OS 2.1, where the error_log was
340 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
341 An intervening stdio call was setting errno as a side-effect, which
342 it has the right to do.
344 *) Erroneous fprintf arguments, a missing return value, an unused
345 function, missing declarations, and other problems detected by gcc
348 *) Duplicated code to create the lock file name merged into a single
349 function, closing several memory leaks.
351 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
352 FCGI app in a given dynamic class would live forever, even if it
353 hadn't been requested for weeks.
355 *) The test for whether to keep looking for dynamic app victims to
356 kill was backwards, so no dynamic apps were ever selected as
359 *) The killInterval and updateInterval were being ignored; the
360 sigsuspend() forced recalculations only when a child died,
361 instead of at the intervals specified.
363 *) When a dynamic app couldn't be started, the program SEGV'd when
364 trying to free the ipcAddrPtr twice.
366 *) Remove the dead lock file and socket when cleaning up after a
367 server whose last child has been killed, so the FastCGIHandler
368 isn't fooled into thinking there is still a process serving
371 *) Make file locking robust in the presence of signals.
373 *) Rename some badly named variables.
375 *) Fix typos in many comments.
377 *) Fix some memory and file descriptor leaks.
379 *) Make the blocking kill of a server closer to working.
381 *) More fixes to calculations in dynamic application management.
384 *** Originally from docs/README.OMI..
386 What's New: Version 2.0b1, 16 Apr 1997
388 *) Implemented a mechanism by which FastCGI applications are started by
389 the web server on the first request and continue running. Also, a
390 heuristics have been implemented to allow for dynamic killing of the
391 running FastCGI apps. The configuration options are supplied via
392 the new FCGIConfig directive, that is described in the
395 *) When performing internal redirect, since the original request's body
396 has already been read, do not allow the redirected request to think
397 that is has one [body].
399 *) More conditional compilation for OS/2.
401 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
403 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
405 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
406 force some output to be written in the period of inactivity. This
407 circumvents the Apache's buffering during the server-push.
409 *) Added a header-parser function placeholder for compatibility with an
412 *) Additional checks have been placed with regard to permissions of the
413 FastCGI processes. These include the existence of the file,
414 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
416 *) Implemented a restart cleanup, so that no parentlesses FastCGI
417 applications are left after Apache is restarted and/or terminated.
418 This was accomplished via two independent mechanisms, where the
419 first one prevents a start of the process manager on the first
420 reading of the configuration files. The second mechanism deals away
421 from Apache's implementation of fork() (via spawn_child()) and
422 implements its own forking, thus removing any dependencies on the
423 Apache to cleanup processes during termination phase.
425 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
426 itself in the file conf.h
428 What's New: Version 1.4.3, 15 Jan 1997
430 *) Fixed compilation warnings for various platforms, as well as
431 conditional compilation for OS/2.
433 What's New: Version 1.4.2, 12 Dec 1996
435 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
436 will proceed under this version of Apache web server.
438 *) As the result of porting, the "include virtual" construct of SSI
439 will now work correctly using either <Location> or <Directory>
442 What's New: Version 1.4.1, 4 Dec 1996
444 *) Checks have been removed from the ScanCGIHeaders that provided
445 for the presence of both Status and Location headers as being
446 an error. Contradictory to CGI/1.1 Internet Draft, both of
447 these headers are used by the current CGI applications.
449 What's New: Version 1.4, 22 Nov 1996
451 *) Added the -port option to AppClass, allowing TCP/IP communication.
452 Added the -socket option to AppClass, allowing Unix domain
453 communication via a configurable pathname.
455 *) Added the ExternalAppClass directive, allowing TCP/IP
456 communication with remote FastCGI applications.
458 *) The handler had its own code for generating HTTP response headers;
459 now it uses Apache's. This reduces the size of the module.
460 More importantly, it fixes the bug in which "include virtual"
461 sees the HTTP response headers.
463 *) The response header parser performed very little checking.
464 Now the parser enforces the guidelines in the CGI/1.1
465 Internet-Draft: Status and Location are mutually exclusive,
466 Location can only be a response to GET or HEAD,
467 CGI response headers can't be repeated, etc. (The CGI response
468 headers are the ones the handler interprets: Status, Location,
471 *) The response header parser used to miss CGI headers
472 with no whitespace after the colon, e.g. "Status:200 OK".
474 *) The response header parser sometimes interpreted the first line
475 of content as an RFC822 continuation line.
477 *) The handler implemented a nonstandard version of Location
478 which never used internal redirects. The handler
479 now attempts to implement Location as specified in the
480 CGI/1.1 Internet-Draft. The module documentation explains
483 *) Error log entries from the response header parser were
484 pretty uninformative; they are better now. If a header is
485 malformed, the log entry includes it. If the headers are
486 unterminated rather than malformed, the log entry says that,
487 and says how many bytes were received from the app.
489 *) When the application manager forked a new process, and that
490 process ran into trouble before executing the first instruction
491 of application code, the process used to exit with status =
492 errno. This made certain configuration problems (e.g. incorrect
493 file permissions when server parent is root) quite difficult
494 to pin down. Now the failing child process opens up the error log
495 and writes an informative entry before exiting.
497 *) The module now correctly handles a slash at the end of the
498 DocumentRoot directive. This was a one-line fix.
500 What's New: Version 1.3.3, 17 Oct 1996
502 *) The module now registers its request handler under the name
503 fastcgi-script in addition to the name application/x-httpd-fcgi.
504 This was a one-line addition to the module, but had quite
505 a large impact on the documentation and sample configuration.
507 What's New: Version 1.3.2, 27 Sept 1996
509 *) On some systems (SunOS, Linux?), fopen for append has a bug
510 that strikes when two processes append to the same file.
511 This bug causes the process manager to corrupt the error log.
512 Work around the bug by calling open, then fdopen. A patch
513 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
514 (Reported by Bob Ramstad.)
516 What's New: Version 1.3.1, 17 Sept 1996
518 *) On Linux and SunOS, the Apache default user_id (-1) is not a
519 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
520 Scott Langley, others.)
522 *) On some systems (Linux, some Solaris?, Irix?), connect requires
523 write access to a Unix Domain socket, so provide it. (Reported
524 by Scott Langley, freeform@wired.com, others.)
526 *) If you hit a .fcg file, but there's no AppClass defined, the
527 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
530 What's New: Version 1.3, 4 Sept 1996
532 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
534 *) Module violated the Apache buff abstraction.
535 Fixed. Might now work with SSL (not tested).
536 As part of the fix, eliminated support for
537 Apache 1.0x versions.
539 *) Module failed to chmod the Unix domain listening sockets
540 it created, so protections were set according to the current
543 *) Module forked too many process manager processes: One
544 per AppClass. When the Apache parent ran as root, these
545 process manager processes ran as root. New process
546 manager is a single process (only started if AppClass is
547 used) and runs with same privileges as other children.
548 The new process manager doesn't do any polling, so
549 there's less system overhead than before.
551 *) AppClass insisted on getting at least two arguments. Fixed.
553 *) Module used setjmp/longjmp, causing compiler warnings
554 on some platforms (e.g. Linux.) The module no longer
557 *) Module created listening sockets in /tmp, where they were
558 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
559 directive to give control over the location of listening
562 *) Module wrote error log entries without a timestamp. Fixed.
564 *) AppClass directive wrote error messages to stderr in addition
565 to returning a char * message to the Apache core. Fixed.
567 What's New: Version 1.2, 3 June 1996
569 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
572 Add version string: APACHEVERSION macro to mod_fastcgi.c
574 *) chown FastCGI socket to user_id and setuid to user_id for app
577 *) Modify GetFromStream() by having it call an OS dependent
578 function GetStreamSize(FILE *) which uses FILE internal data member.
579 Linux users might need to modify GetStreamSize().
581 What's New: Version 1.1, 10 May 1996
583 *) If you specify a non-existent executable in the AppClass directive,
584 or if the file exists and it does not have execute permission you
585 get a constant stream of error messages telling you that "program
586 terminated due to a signal".
588 *) The mod_fastcgi module should use the standard Apache error logging
589 facility instead of writing to stderr.
591 Version 1.0, 30 April 1996