Re: cvs commit: apr/network_io/unix sendrecv.c

[Greg Ames <gr...@apache.org>]

Justin Erenkrantz wrote:
> 
> > gregames@apache.org wrote:

> > While testing this on daedalus, I noticed yet another sendfile_it_all assert
> > dump in /tmp, dated May 10.  I think I'll put a new httpd build into production
> > this weekend with this patch on top of 2.0.36.
> 
> Why don't we roll out with HEAD?  This would get the byte-range
> requests returning 416 on dynamic requests fixed.  -- justin

[gregames@daedalus apache2.0.37dev]$ grep " 416 " /logs/www/weblog | wc -l
     198

OK, you talked me into it, but...

[Mon May 20 07:14:14 2002] [crit] (22)Invalid argument: mod_rewrite: Could not
set permissions on rewrite_log_lock; check User and Group directives
Configuration Failed

(sigh)

Greg

HEAD dies on FreeBSD w/mod_rewrite

[Greg Ames <gr...@apache.org>]

Greg Ames wrote:

> [Mon May 20 07:14:14 2002] [crit] (22)Invalid argument: mod_rewrite: Could not
> set permissions on rewrite_log_lock; check User and Group directives
> Configuration Failed

from apr.h:

#define APR_USE_FLOCK_SERIALIZE           1
#define APR_USE_SYSVSEM_SERIALIZE         0

(gdb) p *rewrite_log_lock.proc_mutex.meth
$13 = {flags = 1, create = 0x280e9cd0 <proc_mutex_flock_create>,
  acquire = 0x280e9d7c <proc_mutex_flock_acquire>, tryacquire = 0,
  release = 0x280e9dd4 <proc_mutex_flock_release>,
  destroy = 0x280e9e2c <proc_mutex_flock_destroy>,
  child_init = 0x280e9e70 <proc_mutex_flock_child_init>}
(gdb) bt
#0  unixd_set_proc_mutex_perms (pmutex=0x81299a8) at unixd.c:410
#1  0x807eaba in unixd_set_global_mutex_perms (gmutex=0x81299a0) at unixd.c:422
#2  0x2820a4da in post_config (p=0x809c010, plog=0x80c6010, ptemp=0x80c8010,
    s=0x80b8118) at mod_rewrite.c:968
#3  0x8065ba4 in ap_run_post_config (pconf=0x809c010, plog=0x80c6010,
    ptemp=0x80c8010, s=0x80b8118) at config.c:129
#4  0x806a973 in main (argc=3, argv=0xbfbffb74) at main.c:579

...so it looks like this lock is an flock, but we're trying to set permissions
on it as if it were a SysV sem. unixd_set_proc_mutex_perms() assumes it is only
called for SysV sems.  I'll take a look at the MPMs to see how they know the
difference.

Greg