| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
NLM is fully qualified.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@64200 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
mod_ssl called apr_os_thread_current() against the 'main' thread we
had not created. Also address the possibility that the pool scope
is bad for a given apr_thread_t and do *not* dereference the ->td
member for apr_os_thread_current().
This patch causes us to 'waste' a system handle for every thread
that *apr* does not create, that apr_os_thread_current() is called
within. In 99% of situations that is a single handle for the main
thread. But there is the possibility of an application creating
dozens of it's own threads outside of apr, each of which then call
apr_os_thread_current(). The scenario appears so abstract and the
complications of this code so obnoxious that this patch has chosen
not to address the possibility.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@64133 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
| |
returning 0xfffffffe (the pseudo-handle) for the current thread.
Stash the real apr_thread_t and recover it for apr_os_thread_current().
We were also missing thread_compare so I dropped that in while I was
at it. The test is simple with the above behavior.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@64085 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@64003 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
"Like the Win32 ExitThread API, _endthreadex does not close the thread handle.
Therefore, when you use _beginthreadex and _endthreadex,
you must close the thread handle by calling the Win32 CloseHandle API."
SUZUKI Rintaro <suzuki@ariel-networks.com> wrote the patch.
Thanks.
- INOUE Seiichiro <inoue@ariel-networks.com>
PR:
Obtained from: SUZUKI Rintaro <suzuki@ariel-networks.com>
Submitted by: INOUE Seiichiro <inoue@ariel-networks.com>
Reviewed by: Will Rowe
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63827 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
passed to sigwait(). It is never valid for them to be there.
Some platforms silently ignore them, some return EINVAL, some
don't process it as desired.
One problem was found with an old Apache 2.0.30 build on AIX.
sig_coredump() wasn't getting called when a plug-in generated SIGABRT.
Removing SIGABRT from the signal mask passed to sigwait() by the
main worker thread fixed the problem. After reviewing sigwait()
documentation it was clear that none of the synchronous signals
should be in the mask passed to sigwait().
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63802 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
(store_desc): Make the backslash quote a newline, not a space.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63725 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
| |
Specifically, True64 with the cc compiler. This resolves that problem
by using a cleaner macro.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63724 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63712 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
| |
these cycles if we aren't toggling USESTDHANDLES or have a handle to
fill-in-the-blanks.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63707 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
| |
CreateProcess() aren't currently initialized properly...
Obtained from: Rob Sacoaccio <robs@fastcgi.com>
Reviewed by: Ian Holsman
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63706 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63591 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
| |
void to apr_status_t. The deprecated versios, apr_*_(un)set_inherit,
are still void, so that we don't inadvertently break code that uses them.
Updated all uses of apr_file_inherit_set in threadproc/win32/proc.c,
and replaced make_handle_private with apr_file_inherit_unset.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63576 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
| |
which reverted 1.76 plus 1.77 ... this code is now restored to 1.77.
This code should now behave as expected, since the include/arch/win32
flavor of inherit.h now makes handles inheritable.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63557 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
| |
The child handles weren't properly inheritable, and redirected command
output got lost in the bit bucket.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63552 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
once again, sorry about that.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63545 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
| |
patches. It seems like extreme overkill to create a pipe when we are
only setting up a log fd for output, and similar cases. We waste an
entire pipe resource just to attach (pipe) to a logfile? That seems goofy.
We now have both _dup and _dup2, so we should take advantage of them.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63544 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63515 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63487 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63472 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
| |
special casing on the inherited bit, but I'm committing for further
comment ... the rest of this is just CE exceptions.
There are no pipes on WinCE 3.0.
Submitted by: Mladen Turk <mturk@mappingsoft.com>
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63466 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
function rather than the NXVmSpawn() stuff.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63439 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63427 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
| |
children from waiting until the parent reaps status, so when the APR
app doesn't care about such status we need to have a handler driven
and call one of the wait functions.
PR: 9168
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63423 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63382 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
| |
apr_signal.h in order to get consistent definitions. In general,
apr_want.h is a bad idea and should be replaced with topic-specific
header files.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63317 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
| |
Some things are easier to follow when you leave them in the natural
order of things.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63314 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
(somehow CVS ignored updating these on the last commit.)
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63287 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
| |
srcdir to always be available in a Makefile, so we need to stop adding
this only when we use VPATH.
Change the dependency generation to use .deps instead of appending to
the Makefile. This makes us consistent with the dependency style of
httpd-2.0.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63286 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
| |
tell APR to ignore SIGCHLD.
Submitted by: Dave Hill <David.D.Hill@Compaq.com>
Reviewed by: Jeff Trawick
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63282 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
_PATH, and APR_SHELLCMD
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63261 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
apr_proc_create so that the parent side of the pipes are opened first.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63250 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
| |
to 5L. The 64-bit definitions for sigset_t changed between
4.X and 5L.
PR: 7957
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63248 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
| |
stop working so hard to figure out the process group id; we
don't use that information (logic was inherited from Apache 1.3,
which does use the information)
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63229 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63220 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63180 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
| |
the cmd.exe path is secure before I'd remove my -1 on beta.
The command.com path gets even more tangled - I don't have the focus
for that this morning.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63179 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
compiling at least once.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63173 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
behaves differently than cmd.exe with respect to the /c command arg.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63172 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
signals which we absolutely must avoid blocking
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63170 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
| |
some debuggers (purify) use it
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63168 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63159 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
"pool".
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63158 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63141 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
spawning an external NLM
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63124 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
| |
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63117 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
| |
progress.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63083 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
| |
to CVS - reverting to 1.63. Although the theory is right - we could
end up in a case of d:\ or \\unc\path\ where stripping off the trailing
slash of a root will introduce erratic behavior or fail entirely.
Needs to be conditional on not-a-root-path, and I don't have time to
hack that tonight. Better to leave working code in CVS. [sorry]
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63074 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
| |
don't know how applications might react. Convention is no trailing
backslash on Win32 cwd paths - so strip it back off when it occurs.
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63073 13f79535-47bb-0310-9956-ffa450edef68
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
forking (aka, to "daemonize").
Detailed explanation: If we are only interested in detaching from the
controlling terminal, then we are only interested in creating a new
process group (or creating a new session, which implicitly creates a
new process group). In order to do so, we must _NOT_ already be
a process group leader. The only way to ensure that is true, we normally
will call fork() and allow the parent to exit, ensuring that the child
is at least a child of a process group leader (and not one itself).
Doing this by default prevents some process-watching tools from working
with Apache. Therefore, when calling apr_proc_detach with
APR_PROC_DETACH_FOREGROUND, the caller is taking responsibility
for _NOT_ being a process group leader, which is guaranteed by
such process management tools.
[A similiar patch was originally submitted Jos and later modifed by Aaron.]
Obtained from: Jos Backus <josb@cncdsl.com>
Submitted by: Aaron Bannert
git-svn-id: http://svn.apache.org/repos/asf/apr/apr/trunk@63044 13f79535-47bb-0310-9956-ffa450edef68
|