Re: Adopt a platform today!

[Richard Frith-Macdonald]

On 22 Feb 2011, at 18:07, Sebastian Reitenbach wrote:

> Hi,
> 
> this happens on OpenBSD/i386 for the base tests:
> 
> 4989 Passed tests
> 140 Failed files
> 108 Failed builds
>   2 Failed tests
>   2 Dashed hopes
> 
> Some of the failed files, about 5 to 10 i killed, since at least some of
> them were running for about half an hour already, so I thought there
> sth. might be screwed? Or is that normal, and I was just too impatient,
> and should rerun it, and leave it running? The tests.sum is attached.

Some tests *are* slow (particularly the I/O related ones where we trying 
reading/writing data in lots of different sized blocks in order to check for 
buffer size issues), and I know you run things on slow machines, but even so 
that sounds like a very long time.
Also, looking at the summary, the output doesn't look right ... so maybe 
there's a problem with the test framework.
I don't know when you checked it out from svn, but there have been a lot of 
changes made in the last few days,  some will have temporarily broken things, 
some were for performance improvements, and some were addressing portability 
issues.  It would be worth checking out a new copy (and making note of the svn 
revision you checked out) and trying that.

It's also possible for individual tests to hang ... I've just found that it 
looks like on solaris 8 the system locking implementation doesn't actually 
support the pthreads api properly and that causes one of the locking tests to 
deadlock :-(