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