Newer
Older
*) Speed up the server's response to a spike in incoming workload
or restarts by assigning empty scoreboard slots to new processes
when they are available. [Greg Ames]
*) Add a handler to mod_includes.c. This handler is designed to
implement the XbitHack directive. This can't be done with a
fixup, because we need to check the content-type, which is
only available in the handler phase. [Ryan Bloom]
*) Make the includes filter check return codes from filters lower in
the filter chain. If a lower level filter returns an error, then
the request needs to stop immediately. This allows mod_include to
stop parsing data once a lower filter recognizes an error.
[Ryan Bloom]
*) Add the ability to extend the methods that Apache understands
and have those methods <limit>able in the httpd.conf. It uses
the same bit mask/shifted offset as the original HTTP methods
such as M_GET or M_POST, but expands the total bits from an int to
an ap_int64_t to handle more bits for new request methods than
an int provides. [Cody Sherr <csherr@covalent.net>]
*) Fix broken mod_mime behavior in merging its arguments. Possible
cause of unexplicable crashes introduced in 2.0.20. [William Rowe]
*) Solve many mod_ssl porting issues (too many to detail) with
help from the whole team, but most notably [Ralf S. Engelschall,
Madhusudan Mathihalli <madhusudan_mathihalli@hp.com>,
Doug MacEachern, William Rowe, Cliff Woolley]
*) More stall fixes for the threaded & worker mpm's.
Make mod_status output more accurate. Don't
count workers in processes which aren't actively
serving requests. [Greg Ames]
*) Win32: Get SSI exec cgi tag working. [Bill Stoddard]
*) Add a single listener/multiple worker MPM. This MPM is
definately not fully correct, but it allows us to solve many
of the problems that exist in the threaded MPM. This is a
modified version of the threaded MPM. [Ryan Bloom]
*) Improve content generation throughout Apache, providing closer
compliance with HTML 3.2, HTML 4.01 Transitional and XHTML 1.0
Transitional specifications. [William Rowe]
*) Fix a problem where the threaded MPM stalls after restarts or
segfaults. Also prevent multiple active processes from using
the same scoreboard slot. [Greg Ames]
*) Apache/Win32 now fills in the service description with Apache's
server version string, including loaded and advertised modules.
[William Rowe]
*) Improved support for the Win32 build, to recover gracefully from
missing apr or apr-util directories or the awk interpreter,
create the proper cgi-bin examples, including a test-cgi.bat, and
fix the perl shebang line for printenv.pl, when installing from
the build environment. [William Rowe]
*) Fix a segfault in threaded.c caused by passing uninitialized
apr_thread_t * to apr_thread_join(). [Jeff Trawick]
*) Use new APR number conversion functions to reduce CPU consumption
when setting the content length, and in mod_log_config.
[Brian Pane]
*) Fix problem reported by Taketo Kabe <kabe@sra-tohoku.co.jp>
where HEAD response headers were being repeated twice for
files greater than 32K bytes (4*AP_MIN_BYTES_TO_WRITE). This
problem in the http_header filter was exposed by the recent rewrite
of the content_length filter. [Taketo Kabe, Bill Stoddard]
*) Fix seg faults in mod_status with ExtendedStatus enabled, after
restarts. A garbage pointer to a vhost's server_rec from the
previous generation was being left around under certain
conditions. [Greg Ames]
*) Fix a cosmetic problem with mod_include. Non-existant SSI vars
used to appear as '(none', without the closing paren.
[Günter Knauf <eflash@gmx.net>]
*) Improve the exports generating awk script. In the past, we had
work around problems in the awk script by avoiding some #if and
#ifdefs. This has bitten us many times in generating the exports.c
file. This improvement allows corrects the header file parsing.
[Sander Striker <striker@apache.org>]
Changes with Apache 2.0.21
*) Resolve the Win32 htpasswd bug, where a file that existed would be
overwritten, regardless of the -c flag.
[William Rowe, Mladen Turk <mladen.turk@mail.inet.hr>]
*) Introduce connection sub-pools into ab. Truncating the lifetime
of these allocations means that ab no longer perpetually grows
its working set, running out of memory on large request attempts.
[William Rowe]
*) Make scoreboard creation a hook. This allows management
modules to have access to the scoreboard at the time that it is
created, and at every restart request.
[Cody Sherr <csherr@covalent.net>]
*) Changed AP_MPMQ_MAX_DAEMONS to refer to MaxClients and
added an AP_MPMQ_MAX_DAEMON_USED to refer to the highest
daemon index actually used in the scoreboard. I also
updated the pertinent calls. [Paul J. Reder]
*) Win32: Prevent listening sockets from being inherited by
the Apache child process, CGI scripts, rotatelog process
etc. If the Apache child process segfaults, any processes
that the child started are not reaped. Prior to this fix,
these processes inherited the listening sockets which sometimes
prevented the restarted Apache child process from accepting
connections (ie, the server would hang).
[Bill Stoddard]
*) Provide vhost and request strings when ExtendedStatus is on.
[Greg Ames]
*) Fix some issues with the pod and prefork: check the pod *after*
processing a connection so that a server processing a time-
consuming request bails out as soon as practical; when the
parent process wakes up a server process via connect(), use an
APR timeout on the connect() so that we don't hang for a long
time if there aren't server processes around to do accept().
[Jeff Trawick, Greg Ames]
*) Performance improvement to mod_mime.c. find_ct() in mod_mime,
spends a lot of time in apr_table_get calls. Using the default
httpd.conf, the tables for languages and charsets are somewhat
large, so the time spent scanning them on each request is
significant. Replacing the tables with hash tables provides
a nice speedup. [Brian Pane <bpane@pacbell.net>]
*) Add two functions to allow modules to access random parts of the
scoreboard. This allows modules compiled for one MPM to access the
scoreboard, even if it the server was compiled for another MPM.
[Harrie Hazewinkel <harrie@covalent.net>]
*) Fix problem in content-length filter where the filter would
buffer all the output from a CGI before sending any bytes
down the filter stack to the network. This problem would cause
significant memory consumption if the CGIs generated
lots of bytes. [Bill Stoddard]
*) Get non-blocking CGI pipe reads working with the bucket brigades.
[Bill Stoddard]
*) Fix seg fault on Windows when serving files cached with mod_file_cache.
[Bill Stoddard]
*) Fix a bug in the threaded MPM that would cause it to kill off all
workers immediately after starting if the number of workers started
was above a certain threshold. [Ryan Bloom, Bill Stoddard]
*) Fix problem with threaded MPM. The problem was that if each child
process was busy serving a single long-lived request and the server
was sent a graceful restart signal, the server would stop serving
requests. This would happen because each child process would wait to
die until the last thread was done, and the parent wouldn't spawn any
new children until a process died. Now, the parent looks at the fact
that the children are dying gracefully, and starts new children.
Those new children only start enough threads to compliment the number
of threads in the other child process that shares the same spot in
the scoreboard. In this way, we make sure to never go over
MaxClients. [Ryan Bloom]
*) modified mod_negotiation and mod_autoindex to speed up by almost a
factor of two on apr_dir_read()-enhanced platforms, such as Win32
and OS2, by calling ap_sub_request_lookup_dirent() with the results
already provided by apr_dir_read(). [William Rowe]
*) mod_file_cache is now more robust to filtering and serves requests
slightly more efficiently. [Cliff Woolley]
*) Fix problem handling FLUSH bucket in the chunked encoding filter.
Module was calling ap_rwrite() followed by ap_rflush() but the
served content was not being displayed in the browser. Inspection
of the output stream revealed that the first data chunk was
missing the trailing CRLF required by the RFC. [Bill Stoddard]
*) apxs no longer generates ap_send_http_header() in the example handler
*) Fix an ab problem which could cause a divide-by-zero exception
with certain invocations (e.g., ab -k -c 6 -n 100 localhost/).
[Ian Holsman <ianh@cnet.com>]
*) Solve case-insensitive platforms' confusion about negotiated
filenames, allowing files of differnt case to match in choosing
the document to serve. [William Rowe]
Loading full blame...