Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Geoffrey Young


Randy Kobes wrote:
On Mon, 17 Nov 2003, Geoffrey Young wrote:


I dunno, but I am afraid that stuff like device, inode,
nlink and such might not work similarly on non-Unixes..
These tests most likely need to be a bit more
OS-specific to avoid failures on those OSes
yes, I meant to follow this up with an email saying that I
would quickly fix any os related issues once randy et al
get the chance to play with them and expose where APR is
different than perl.  but I didn't want to just skip them
on win32 before I knew where the issues were, and testing
checked in stuff is just easier :)


That's true ... On Win32, inode, device, user, and group are
all defined, but these tests fail as written, as the
returned values are different ... The same thing goes for
the test involving WEXECUTE. Probably one might come up with
what these returned values would be on Win32, but I'm
wondering if it's worth it - from what I gather, some of
these might depend on the Win32 flavour being used (eg, on
Win9?, any readable file is [almost always] executable).
Perhaps we could just skip these tests on Win32:
if the semantics are really that dissimilar, I guess it makes the most sense 
to just skip them, especially since the concept of user and group aren't 
really meaningful on win32.  I'm glad to see the RW stuff work, though :)

your patch looks good to me, so commit at will.

--Geoff

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Randy Kobes
On Tue, 18 Nov 2003, Geoffrey Young wrote:

> Randy Kobes wrote:
> > On Mon, 17 Nov 2003, Geoffrey Young wrote:
> >
> >>>I dunno, but I am afraid that stuff like device, inode,
> >>>nlink and such might not work similarly on non-Unixes..
> >>>These tests most likely need to be a bit more
> >>>OS-specific to avoid failures on those OSes
> >>
> >>yes, I meant to follow this up with an email saying that I
> >>would quickly fix any os related issues once randy et al
> >>get the chance to play with them and expose where APR is
> >>different than perl.  but I didn't want to just skip them
> >>on win32 before I knew where the issues were, and testing
> >>checked in stuff is just easier :)
> >
> > That's true ... On Win32, inode, device, user, and group are
> > all defined, but these tests fail as written, as the
> > returned values are different ... The same thing goes for
> > the test involving WEXECUTE. Probably one might come up with
> > what these returned values would be on Win32, but I'm
> > wondering if it's worth it - from what I gather, some of
> > these might depend on the Win32 flavour being used (eg, on
> > Win9?, any readable file is [almost always] executable).
> > Perhaps we could just skip these tests on Win32:
>
> if the semantics are really that dissimilar, I guess it
> makes the most sense to just skip them, especially since
> the concept of user and group aren't really meaningful on
> win32.  I'm glad to see the RW stuff work, though :)
>
> your patch looks good to me, so commit at will.

OK, I'll do that - thanks ... It is great that the others
(RW, size, times) do work. I think the thing with the
user/group, indode/device, and execute is that, since
there's no universally obvious way to deal with these on
Win32, APR does one thing and Perl does another, and the two
don't always agree with each other (and also, as I indicated
before, the specific disagreement may also depend on the
Win32 flavour).

However, the fact that the $r->finfo->fname returns undef on
Win32 is something that should be tracked down. At the
moment I don't see where the trouble lies, as it's statting
the right file; it's just that the name isn't getting
recorded.

-- 
best regards,
randy

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Geoffrey Young

However, the fact that the $r->finfo->fname returns undef on
Win32 is something that should be tracked down. At the
moment I don't see where the trouble lies, as it's statting
the right file; it's just that the name isn't getting
recorded.
oh, I missed that - yes, that ought to be tracked down.  are you on apr@ and 
can post it there?  if not, I'll see if I can't track down bill rowe here 
(at apachecon) and talk to him about it.

--Geoff

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Geoffrey Young


Geoffrey Young wrote:

However, the fact that the $r->finfo->fname returns undef on
Win32 is something that should be tracked down. At the
moment I don't see where the trouble lies, as it's statting
the right file; it's just that the name isn't getting
recorded.


oh, I missed that - yes, that ought to be tracked down.  are you on apr@ 
and can post it there?  if not, I'll see if I can't track down bill rowe 
here (at apachecon) and talk to him about it.
actually, I see where the problem is, so I'll post it and report back.

--Geoff

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Randy Kobes
On Tue, 18 Nov 2003, Geoffrey Young wrote:

> > However, the fact that the $r->finfo->fname returns undef on
> > Win32 is something that should be tracked down. At the
> > moment I don't see where the trouble lies, as it's statting
> > the right file; it's just that the name isn't getting
> > recorded.
>
> oh, I missed that - yes, that ought to be tracked down.
> are you on apr@ and can post it there?  if not, I'll see
> if I can't track down bill rowe here (at apachecon) and
> talk to him about it.

No, I'm not on apr@ ... And actually, I'm not sure if it's a
problem on the Perl side or the apr side ...  I can't see
what can go wrong on the Perl side, though, given, for
example, $r->finfo gets the size of the file correct. But
stranger things have happened ... If you do see bill rowe,
it would be good to know if he knows about this - there was
a post on the apr-dev list on 2003-01-10 (that he responded
to) about a patch to set both info.name and info.fname on
Win32, but it looks like that there were some problems with
the patch.

-- 
best regards,
randy

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Re: mod_perl segfaults when also using mod_rewrite

2003-11-18 Thread Aaron Paetznick
When I do that, the port doesn't even seem to open.  Here's what I do:

[EMAIL PROTECTED] mod_perl-1.99_11]# gdb /usr/local/www/apache-2.0.48/bin/httpd
GNU gdb Red Hat Linux (5.2.1-4)
Copyright 2002 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain 
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i386-redhat-linux"...
(gdb) run -X -d /usr/local/www/apache-2.0.48
Starting program: /usr/local/www/apache-2.0.48/bin/httpd -X -d 
/usr/local/www/apache-2.0.48
[New Thread 16384 (LWP 28385)]

...and I can't even attach to the port then.  Any ideas?

--Aaron



Bobber Cheng wrote:
1. compile with debug flag
First make sure u install gdb, then compile httpd and modperl in debug mod,
configure apache2 as:
./configure --enable-maintainer-mode and ur custom flages
configure mod_perl ad:
perl "Makefile.PL  MP_DEBUG=1 and ur custom flages"
compile and install them to the right place.
2. run apache2 in gdb
$APACHE2_HOME is ur apache installed directory such as /usr/local/apache2
then run gdb as:
gdb $APACHE2_HOME/bin/httpd
(gdb) run -X -d $APACHE2_HOME
apache2 should run.

3. report the backtrace
visit ur page us usally, when server segfaults, gdb will tell ur where. 
u just type
(gdb) backtrace
and report the result.

hope it could help u.

[EMAIL PROTECTED] wrote:

I'm using apache v2.0.48 and have confirmed that the problem exists 
with both mod_perl v1.99_11 and v1.99_12-dev in CVS.  If I simply use 
"RewriteEngine On", mod_perl works fine.  If I use "RewriteEngine 
OnRewriteRule ^/(.*) /index.pl/$1 [PT]" (or anything similar), the 
apache server immediately segfaults, displays nothing, and logs 
nothing useful (even in debug).  This does not happen with mod_perl 
v1.99_10, although for reasons unrelated to mod_perl I'm actually 
stuck using v1.99_09 right now, which works more or less fine with 
this configuration.  I'm using apache compiled from source on RedHat 
8.0, with perl-5.8.0-88.3 and GCC v3.3.2.

Unfortunately I'm not a C programmer, or I probably would have been 
able to send more useful information.  I tried to follow the 
instructions on debugging mod_perl C internals and get a core file, 
but I couldn't get that to work either.  Any ideas?  Thanks.

--Aaron





-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Geoffrey Young
If you do see bill rowe,
it would be good to know if he knows about this - there was
a post on the apr-dev list on 2003-01-10 (that he responded
to) about a patch to set both info.name and info.fname on
Win32, but it looks like that there were some problems with
the patch.
I just spent 30 minutes with him on this and it seems that it's not that 
complex, but there are a few caveats that need to be noted.  nonetheless, he 
said to watch apr space over the next day or so for a fix :)

so, as always, thanks randy!

--Geoff

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: mod_perl segfaults when also using mod_rewrite

2003-11-18 Thread Stas Bekman
Aaron Paetznick wrote:
When I do that, the port doesn't even seem to open.  Here's what I do:

[EMAIL PROTECTED] mod_perl-1.99_11]# gdb /usr/local/www/apache-2.0.48/bin/httpd
GNU gdb Red Hat Linux (5.2.1-4)
Copyright 2002 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you 
are
welcome to change it and/or distribute copies of it under certain 
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i386-redhat-linux"...
(gdb) run -X -d /usr/local/www/apache-2.0.48
Starting program: /usr/local/www/apache-2.0.48/bin/httpd -X -d 
/usr/local/www/apache-2.0.48
[New Thread 16384 (LWP 28385)]

...and I can't even attach to the port then.  Any ideas?
port? Did you mean thread? If you are running a prefork mp you should do:

gdb> run -DONE_PROCESS -DNO_DETATCH

If a worker/winnt mpm, then:

gdb> run -DONE_PROCESS -DNO_DETATCH -DPERL_USEITHREADS

For the complete info please read:
http://perl.apache.org/docs/2.0/devel/debug/c.html#Debug_notes
__
Stas BekmanJAm_pH --> Just Another mod_perl Hacker
http://stason.org/ mod_perl Guide ---> http://perl.apache.org
mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com
http://modperlbook.org http://apache.org   http://ticketmaster.com
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: mod_perl segfaults when also using mod_rewrite

2003-11-18 Thread Stas Bekman
Stas Bekman wrote:

gdb> run -DONE_PROCESS -DNO_DETATCH

If a worker/winnt mpm, then:

gdb> run -DONE_PROCESS -DNO_DETATCH -DPERL_USEITHREADS
I need to learn to spell, please

perl -pi -e 's|DETATCH|DETACH|g' << THIS_MESSAGE

__
Stas BekmanJAm_pH --> Just Another mod_perl Hacker
http://stason.org/ mod_perl Guide ---> http://perl.apache.org
mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com
http://modperlbook.org http://apache.org   http://ticketmaster.com
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: cvs commit: modperl-2.0/t/response/TestAPR finfo.pm

2003-11-18 Thread Randy Kobes
On Tue, 18 Nov 2003, Geoffrey Young wrote:

> > If you do see bill rowe,
> > it would be good to know if he knows about this - there was
> > a post on the apr-dev list on 2003-01-10 (that he responded
> > to) about a patch to set both info.name and info.fname on
> > Win32, but it looks like that there were some problems with
> > the patch.
>
> I just spent 30 minutes with him on this and it seems that
> it's not that complex, but there are a few caveats that
> need to be noted.  nonetheless, he said to watch apr space
> over the next day or so for a fix :)
>
> so, as always, thanks randy!
>
> --Geoff

Great  So for now, should we skip the fname test on
Win32 until a future Apache2 version comes with the patch?

-- 
best regards,
randy

-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



[Fwd: mod_perl-1.99_11 on HP-UX 11.22]

2003-11-18 Thread Stas Bekman
forwarding this to where this message belongs

 Original Message 
Subject: mod_perl-1.99_11 on HP-UX 11.22
Date: Tue, 18 Nov 2003 15:16:49 -0800 (PST)
From: TJ Saunders <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Hello, Stas.  I'm building mod_perl-1.99_11 using Apache 2.0.48 and perl
5.8.2 on HP-UX 11.22 using HP's C compiler -- everything built just fine.
However, when I went to run the mod_perl test suite, quite a few of the
tests failed.  The error in the t/logs/error_log was the "Can't locate foo
in @INC (...)" message, which unfortunately is not verbose enough to
pinpoint the culprit.
Using tusc, I eventually tracked it down to open() returning EMFILE for
some of the test suite's require()s.  Turns out the default number of
per-process open files in the shell was quite low: 60.  Setting 'ulimit -n
1024' made the tests pass successfully.
If possible, could this checking/setting of this resource limit be done by
TestRun.pm, similarly to what is done for the core size limit, so that it
doesn't bite other developers?
Cheers,
TJ
---
 TJ <[EMAIL PROTECTED]>
---
--

__
Stas BekmanJAm_pH --> Just Another mod_perl Hacker
http://stason.org/ mod_perl Guide ---> http://perl.apache.org
mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com
http://modperlbook.org http://apache.org   http://ticketmaster.com
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: [Fwd: mod_perl-1.99_11 on HP-UX 11.22]

2003-11-18 Thread Stas Bekman
Stas Bekman wrote:
forwarding this to where this message belongs

 Original Message 
Subject: mod_perl-1.99_11 on HP-UX 11.22
Date: Tue, 18 Nov 2003 15:16:49 -0800 (PST)
From: TJ Saunders <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Hello, Stas.  I'm building mod_perl-1.99_11 using Apache 2.0.48 and perl
5.8.2 on HP-UX 11.22 using HP's C compiler -- everything built just fine.
However, when I went to run the mod_perl test suite, quite a few of the
tests failed.  The error in the t/logs/error_log was the "Can't locate foo
in @INC (...)" message, which unfortunately is not verbose enough to
pinpoint the culprit.
Using tusc, I eventually tracked it down to open() returning EMFILE for
some of the test suite's require()s.  Turns out the default number of
per-process open files in the shell was quite low: 60.  Setting 'ulimit -n
1024' made the tests pass successfully.
If possible, could this checking/setting of this resource limit be done by
TestRun.pm, similarly to what is done for the core size limit, so that it
doesn't bite other developers?
Thanks for tracing this problem, TJ. I guess we could do what you propose (as 
we already use ulimit to allow core files), though I'm not entirely sure that 
it's safe to hardcode it as 1024. What if the system's hard limit is set to a 
lower number? What if someone else's test suite will need to open more than 
1024 files? (since Apache-Test is used by many other projects)

Therefore I think we should set it to one of the following: hard,  soft,  or 
unlimited. Probably "hard" is the best choice. I'm not sure how cross-platform 
this is, but that's what my bash manpage on linux suggests.  Though it doesn't 
see to work for -n.

Of course one could check /proc/sys/fs/file-nr but again I don't think this is 
cross-platform.

Further suggestions are very welcome.

__
Stas BekmanJAm_pH --> Just Another mod_perl Hacker
http://stason.org/ mod_perl Guide ---> http://perl.apache.org
mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com
http://modperlbook.org http://apache.org   http://ticketmaster.com
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Re: [mp2] segfaults when using "include" filter w/mod_perl

2003-11-18 Thread Stas Bekman
Tyler 'Crackerjack' MacDonald wrote:
Hello,

 Thanks for taking the time to look at this:
Thanks for reporting this Tyler, I'm looking at this issue, though I may come 
back to you only on Thu, once I'm back from ApacheCon.

__
Stas BekmanJAm_pH --> Just Another mod_perl Hacker
http://stason.org/ mod_perl Guide ---> http://perl.apache.org
mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com
http://modperlbook.org http://apache.org   http://ticketmaster.com
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]