Re: default branch test failures (22 ouch!)

[Ben Abbott]

On Oct 1, 2011, at 12:31 PM, Ben Abbott wrote:

> On Sep 30, 2011, at 3:23 PM, Ben Abbott wrote:
> 
>> On Sep 30, 2011, at 10:44 AM, Martin Helm wrote:
>> 
>>> Am Freitag, den 30.09.2011, 08:20 -0400 schrieb Ben Abbott: 
>>>> I'm seeing failures for 
>>>> 
>>>> scripts/statistics/base/corrcoef.m ..................... PASS 6/8 FAIL 2 
>>>> test_classes.m ......................................... PASS   21/37   
>>>> FAIL 16
>>>> test_ctor_vs_method.m .................................. PASS    1/4    
>>>> FAIL 3
>>>> test_string.m .......................................... PASS   78/79   
>>>> FAIL 1
>>>> 
>>>> Its been a week or so since I ran "make check". Last time I did there were 
>>>> no failures. Is anyone else seeing these?
>>>> 
>>>> I've attached a zipped fntests.log
>>>> 
>>>> Ben
>>>> 
>>> 
>>> Cannot see that, just refreshed and rebuilt but made it clean before and
>>> ran the full thing (autogen.sh -> configure). Did you try with a make
>>> clean?
>>> 
>>> Summary:
>>> 
>>> PASS   9692
>>> FAIL      0
>>> 
>>> There were 2 expected failures (see fntests.log for details).
>> 
>> I did a "make maintainer-clean", autogen, configure, and "make check" but 
>> still see the 22 errors.
>> 
>> I'll revert to a week or two ago and verify all tests pass. If they do not, 
>> then something may have changed due to updates via my package manager. If I 
>> can return to all tests passing, I'll use "hg bisect" to isolate the problem.
>> 
>> Ben 
> 
> I reverted 7 days back, but still encountered the errors. However, since the 
> stable branch doesn't passes the tests, I'm concluding I just need to revert 
> back further.
> 
> Ben

I resolved the problem.

I had tried "make maintainer-clean" followed by autogen, configure, and then 
make but the errors persisted. Today I began with ...

        hg status | grep '^? ' | sed "s/^? /rm /g" | /bin/sh 

Which I had thought was equivalent to a maintainer-clean. However, this did the 
trick.

I assume there is something about the build process that isn't always robust on 
MacOS. Might it have to do with have a case insensitive file system?

Ben