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