Aaron Bannert [Fri, 19 Oct 2001 23:36:55 +0000 (23:36 +0000)]
The worker MPM is now completely ported to APR's new lock API. It
uses native APR types for thread mutexes, cross-process mutexes,
and condition variables. Who's next?
Aaron Bannert [Fri, 19 Oct 2001 23:32:43 +0000 (23:32 +0000)]
Use the APR's new OS-specific proc mutex accessors -- they are used
here to set permissions on SysV Semaphores. MPMs will be modified to
call this new function as they are ported to the new APR lock API.
Bill Stoddard [Fri, 19 Oct 2001 21:11:03 +0000 (21:11 +0000)]
Win32: ap_max_requests_per_child should not be a local static in mpm_winnt.c.
This was preventing the MaxRequestsPerChild config directive from working
properly.
Handle MaxRequestsPerChild a bit more sanely. When MaxRequestsPerChild is hit,
the child signals an event that causes the main child thread to signal the parent
to restart.
Hey, we tagged 2.0.26. =)
We're now waiting for it to be discarded or we need to bump the tag on
server/core.c. And, we need to fix the proc_wait thing to someone's
satisfaction.
Remove all references to the Port directive and replace it with
appropriate links to the Listen directive.
If I had known that so many files had the Port directive in it, I would
have bundled this up as one big commit. *sigh* I just kept looking
and finding new files, fixing, and committing. So, this should be the
rest of them for now.
Aaron Bannert [Wed, 17 Oct 2001 21:34:29 +0000 (21:34 +0000)]
Gets VPATH builds working again. The problem was that apr.h is generated
by autoconf and then placed in the builddir, not the srcdir (which we
didn't notice until we tried to do VPATH). When cpp couldn't find apr.h
we would decide not to build mod_so, which is kinda bad.
Now we include both paths when we test for APR_HAS_DSO and everyone is happy.
Aaron Bannert [Wed, 17 Oct 2001 16:29:37 +0000 (16:29 +0000)]
Fully convert worker's fdqueue to the new APR lock API:
- Mutexes and Condition Variables from APR are now used instead of
direct (non-portable) calls to pthread.
Still need to fully APRize these routines, particularly the return
values. In many cases a failed lock or cond var call will simply return
FD_QUEUE_FAILURE. This should obviously instead return the appropriate
APR error code.
Joshua Slive [Wed, 17 Oct 2001 15:57:04 +0000 (15:57 +0000)]
I certainly see the advantage of having SSL split out into another file.
And indeed, that is the way that I would do it for my own server. But
the default config file is really there for beginners, not for advanced
users. Having multiple config files is bad for beginners because it
makes it difficult to answer questions like "where do I put this config
directive?" It also makes it more difficult to move a specific configuration
from one server to another. Finally, it is just overall more intimidating
to be faced with a bunch of files to edit, rather than one file that
controls everything.
Having said that, it does look like the overall feeling is that at least
SSL should be split out, and I won't object (strongly).
Bill Stoddard [Wed, 17 Oct 2001 15:51:22 +0000 (15:51 +0000)]
Win32: Deprecate ap_start_shutdown/ap_start_restart in favor of ap_signal_parent(enum).
This simplifies the code a bit for some more MPM cleanup work forthcoming.
Jim Jagielski [Wed, 17 Oct 2001 15:32:22 +0000 (15:32 +0000)]
Not sure if the number of directives in a normal proxy setup would
warrant needing a whole separate config file. I see such break outs
as pretty much dependent on the number of lines it adds to
httpd.conf... with SSL it makes sense. With proxy, well, I'm not
sure. To be honest, though, breaking out both won't trouble me
a bit, but this is my pref.
PR:
Obtained from:
Submitted by:
Reviewed by:
Ken Coar [Wed, 17 Oct 2001 14:50:48 +0000 (14:50 +0000)]
The argument that this is a reversion to the httpd/access/srm
multifile scenario is bogus, IMHO; the problem with those
was that their purposes were not clear, nor were the orders
of invocation. Putting massive (or tiny but uncommon) chunks
of easily self-contained special-function directives into
separate files simplifies the basic configuration and doesn't
revert to either of the trinity-file complaints, since these
would need to be explicitly invoked with an Include directive.
Aaron Bannert [Wed, 17 Oct 2001 00:59:18 +0000 (00:59 +0000)]
First pass at converting worker MPM to APR's new lock API. This converts
the APR_INTRAPROCESS locks into lightweight, non-nested apr_thread_mutex_t
locks.
It is absolutely invalid practice to test 'prot' bits to determine if a
file is readable. The only acceptable means of testing readability is to
open it for reading, due to discrepancies between permissions, DACLs and
SACLS. Even Linux hackers are gonna need to learn that lesson if they
plan to do any DOD or Gov work once DACL-enhanced Linux is adopted.
Ryan Bloom [Tue, 16 Oct 2001 21:32:41 +0000 (21:32 +0000)]
Remove all of the ldap modules. These have been moved to their own
repository, httpd-ldap, and they now form their own sub-project of the
httpd project.
LoadFile(s) required by modules before they loaded weren't executed until
server initialization. This patch assures all such loaded files occur as
they are parsed.
Submitted by: John Sterling <sterling@covalent.net>
Jeff Trawick [Tue, 16 Oct 2001 18:45:16 +0000 (18:45 +0000)]
Exit when we can't listen on any of the configured ports. This
is the same behavior as 1.3, and it avoids having the MPMs to
deal with bogus ap_listen_rec structures.
This also backs out some circumventions I and Greg Ames had added
to prefork; these are no longer necessary because of this change.
Aaron Bannert [Tue, 16 Oct 2001 17:51:11 +0000 (17:51 +0000)]
In one target we weren't always creating the $libexecdir
($prefix/modules) directory, and in another we were assuming that the
directory existed. Now we always create it in the first target, and we
check that it exists in the second.
This fixes a bug where httpd.exp would be installed as a _file_ called
$prefix/modules.
Ryan Bloom [Tue, 16 Oct 2001 05:18:39 +0000 (05:18 +0000)]
Cleanup the proxy code that creates a request to the origin
server. This change adds an optional hook, which allows modules
to gain control while the request is created if the proxy module
is loaded. The purpose of this hook is to allow modules to add
input and/or output filters to the request to the origin. While
I was at it, I made the core use this hook, so that proxy request
creation uses some of the code from the core. This can still be
greatly improved, but this is a good start.
Streamline this function, and append the trailing slash for any directories
that are resolved. Needs to be more tightly coupled to the dir_walk
optimized cache.
This will probably fix recent breakage to mod_negotation and httpd.test,
as well as the /manual/ returning docroot/index. Need to look for another
solution. I'm suspecting path_info is possibly broken.
Aaron Bannert [Tue, 16 Oct 2001 01:04:08 +0000 (01:04 +0000)]
Cautiously avoiding the scoping issue that was discussed on the list,
I thought these other changes needed to go in; Namely that we don't
need to check if the brigade is empty twice in the loop, just once.
Added a comment for good measure.
Doug MacEachern [Mon, 15 Oct 2001 23:14:11 +0000 (23:14 +0000)]
disabling mod_example ap_hook_{http_method,default_port}
breaks things such as httpd-test when http_method is "foo"
PR:
Obtained from:
Submitted by:
Reviewed by:
Clarify the new MultiviewsMatch directive. I wanted this down on 'paper',
even if it isn't proofread - please feel free to touch up.
BTW - the new 'cleaned' formatting really stinks for hand-editing,
the indentation really needs works. Our per-directive meta-info is
the ugliest aspect. Would be good if the cleanup code could be customized
to make the html not only clean, but legible ;)
Fix the first bucket not getting chunked properly. The reason was that
we left the TE header in the output headers, so ap_set_keepalive didn't
set r->chunked to 1. So, ap_http_header_filter wouldn't insert CHUNK
filter appropriately. I have no clue how it chunked anything at all
before this patch, but hey, it looks right now.
Tested with Yahoo, Apple, apache.org, and /.
/. is interesting in that it sends both Conn: Close and TE: chunked.
Well, now I know what the bio_is_renegotiating call was for.
Place a big-ass comment there so that whomever comes next isn't stuck
at a cryptic call that they don't understand with a dinky comment.
Hopefully, this makes sense. Someone more familiar with OpenSSL should
verify the comment.
This fix also requires the normalize call to be performed before
churn_input so that we don't enter churn_input with a 0-length ctx->b
brigade.
All httpd-test tests (except for the module/negotiation test) pass now.
projects / apache / log