Re: svn 1.2.1 os x: make check fails b/c of too many open files

Apparently, I left out an important piece of the puzzle (oops).

I'm building against Apache 2.1.6-alpha.  When I remove the "--with- 
apr=/usr/local/apache2/bin/apr-1-config --with-apr-util=/usr/local/ 
apache2/bin/apu-1-config" part from my configure script, the tests pass.

Anybody else tried this configuration?

stephen

On Aug 8, 2005, at 2:08 PM, Karan, Cem (Civ, ARL/CISD) wrote:

> I'm running on OS X 10.4.2, and I just recompiled using ./ 
> configure; make; make check, and I'm NOT having any of these  
> problems....
>
> I just did a quick hack around to see what I can do with the ulimit  
> before I start causing failures.  It looks like somewhere in the  
> range of 25-50 for me.  As a question to the original poster, are  
> you using a dual processor machine?  I know that there have been a  
> LOT of problems reported for dual processor machines with XCode  
> (mostly with distcc chewing up 70%+ of both CPUs) but there may  
> also be problems where something down in the kernel is not properly  
> closing files.
>
> Thanks,
> Cem Karan
>
> -----Original Message-----
> From: Philip Martin [mailto:[email protected]]
> Sent: Sun 07-Aug-05 05:48 PM
> To: [email protected]; [email protected]
> Subject: Re: svn 1.2.1 os x: make check fails b/c of too many open  
> files
>
> [email protected] writes:
>
>
>> Stephen, I'm passing your mail along to dev@.  If this is
>> reproducable, it's a real problem.  Folks, see below:
>>
>
> I think it must be an OSX thing, I can't reproduce it on my Linux box.
>
>
>>> START: diff-diff3-test
>>> PASS:  diff-diff3-test 1: these dump core
>>> svn_tests: Can't open file 'bar10': Too many open files
>>> FAIL:  diff-diff3-test 2: 2-way unified diff and trivial merge
>>>
>
> This test runs 17 diffs one after the other, it doesn't use an explict
> subpool.  'bar10' is one of the files in the 10th diff, so it appears
> to have run 9 diffs before running out of file handles.
>
>
>>> PASS:  diff-diff3-test 3: 2-way unified diff where output is suspect
>>>
>
> main() clears the test pool before running each test.  That appears to
> be enough to make test 3 work after test 2 above fails
>
>
>>> PASS:  diff-diff3-test 4: 3-way merge, non-overlapping changes
>>> PASS:  diff-diff3-test 5: 3-way merge, non-conflicting overlapping
>>> changes
>>> PASS:  diff-diff3-test 6: 3-way merge, conflicting overlapping  
>>> changes
>>> PASS:  diff-diff3-test 7: random trivial merge (seed:3366488657)
>>> PASS:  diff-diff3-test 8: random 3-way merge (seed:3461334650)
>>> PASS:  diff-diff3-test 9: merge with part already present (seed:
>>> 1461195583)
>>> PASS:  diff-diff3-test 10: 3-way merge, adjacent changes
>>> END: diff-diff3-test
>>>
>
> On my linux box I can do 'ulimit -n 7' (max 7 open files) and run
> diff-diff3-test without a problem.  If I do 'ulimit -n 6' all the
> tests fail.  In particular the second test fails on the first of the
> 17 diffs, not on the tenth ('foo4' rather than 'bar10'):
>
> svn_tests: Can't open file 'foo4': Too many open files
> FAIL:  lt-diff-diff3-test 2: 2-way unified diff and trivial merge
>
> We need someone with an OSX box to debug this.
>
> -- 
> Philip Martin
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>