[Gluster-devel] Problems with debugging NetBSD failures.

[Ravishankar N]

[rant]

Hi,

We are observing many regression tests that fail only on NetBSD. Some of 
them genuine and expose problems in the patch while many are spurious. 
While I'm not saying that we do away with NetBSD, it is extremely 
painful to set up the jenkins slaves to do any meaningful debugging. I 
wasn't able to find standard stuff like cscope, vim and strace  on the 
jenkins machines. gdb was segfaulting as soon as the first breakpoint 
was hit. The core did not have any debug symbols despite building with 
--enable-debug. The .t files do not complete if you comment out the 
cleanup in the end and run it via prove. Learning equivalent tools 
(ktrace et.al) is time consuming and installing cscope etc. was impossible.

If we are investing time and effort in solving NetBSD failures (despite 
the fact that *downstream* does not care about NetBSD), it would be 
necessary to at least have a VM image of NetBSD with all development 
tools installed so that we can run it on our linux laptops. Otherwise I 
can easily see run_tests.sh being the single file in gluster repo having 
the most number of commits because we keep adding and removing tests 
to/from is_bad_test(). Is there a good soul who wants to volunteer to 
build a NetBSD VM image that can run on linux? Or does anyone have a 
better idea?

Regards,
Ravi