[Gluster-devel] is tests/basic/gfapi/libgfapi-fini-hang.t broken in NetBSD ?

[Niels de Vos]

On Fri, Jul 07, 2017 at 07:42:08AM -0400, Jeff Darcy wrote:
> 
> 
> On Fri, Jul 7, 2017, at 03:36 AM, Niels de Vos wrote:
> > The segfault is caused by GF_ASSERT() on
> > https://review.gluster.org/#/c/17662/2/libglusterfs/src/mem-pool.c@563 .
> > At the moment I'm not sure how this can happen, unless glfs_fini() is
> > called more than once on a glfs_t object.
> 
> It's because the test deliberately calls glfs_fini without calling
> glfs_init.  Unlike most APIs, in GFAPI "init" is neither a module-level
> operation  nor the first thing users should call.  Looking at what
> already happens in glfs_new vs. glfs_fini, it looks like the call to
> mem_pools_init should have gone into glfs_new anyway.  It's easy to fix;
> I'll send a patch for it shortly.

Thanks! Yes, glfs_new() would probably be more appropriate.

> The other mystery is why this same test passed on Centos.  Seems like it
> should have failed in exactly the same way.

A difference could be that ./configure is called with --enable-debug on
NetBSD, but not on CentOS. I think GF_ASSERT() only logs when in
non-debug mode.

Niels
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://lists.gluster.org/pipermail/gluster-devel/attachments/20170707/b178333f/attachment.sig>