Re: [Qemu-devel] test-qga failures on Fedora 24

[Michael Roth <mdroth@linux.vnet.ibm.com>]

Quoting John Snow (2016-07-06 17:41:32)
> 
> 
> On 07/06/2016 06:20 PM, Michael Roth wrote:
> > Quoting Marc-André Lureau (2016-07-06 16:40:52)
> >> Hi
> >>
> >> ----- Original Message -----
> >>> I upgraded to Fedora 24 last Friday before the long (American) weekend.
> >>> Probably a bad idea.
> >>>
> >>> Regardless, test-qga has developed a new symptom on current development
> >>> HEAD that I don't know what to make of:
> >>>
> >>> https://paste.fedoraproject.org/388356/14678205/
> >>>
> >>> Here's the meat of it:
> >>>
> >>>   LINK  tests/test-qga
> >>> GTESTER tests/test-qga
> >>> **
> >>> ERROR:/home/bos/jhuston/src/qemu/tests/test-qga.c:266:test_qga_get_fsinfo:
> >>> assertion failed ret: GenericError realpath("/sys/dev/block/8:17"): No
> >>> such file or directory
> >>> GTester: last random seed: R02S9d55aa8939b9bd797279bd3f69d33466
> >>> **
> >>> ERROR:/home/bos/jhuston/src/qemu/tests/test-qga.c:685:test_qga_blacklist:
> >>> assertion failed ret: GenericError realpath("/sys/dev/block/8:17"): No
> >>> such file or directory
> >>> GTester: last random seed: R02Sb9bc461c2129c284f83033dca27827c6
> >>> /home/bos/jhuston/src/qemu/tests/Makefile.include:684: recipe for target
> >>> 'check-tests/test-qga' failed
> >>> make: *** [check-tests/test-qga] Error 1
> >>>
> >>>
> >>> Any clues as to what's gone awry?
> >>>
> >>
> >>
> >> build_guest_fsinfo_for_device() failed, it's probably worth to investigate why the agent treats this as a fatal error, this would likely happen in a guest too. Could you try to reproduce in a VM or give clues on your setup to reproduce it?
> >>
> > 
> > The code assumes (non-virtual) block devices have a symlink at
> > /sys/dev/block/<major>:<minor>, maybe this isn't the case in Fedora
> > anymore?
> > 
> > The alternative is that realpath("/sys/dev/block/8:17", NULL) is what's actually
> > failing. That call assumes the first argument is a symlink, maybe that's
> > not the case on FC24?
> > 
> > Don't have an fc24 handy, but can check tomorrow.
> > 
> > 
> 
> Good news: The problem went away.
> 
> It was likely due to an external XFS drive I have that perhaps didn't
> properly register as ejected. I unmounted it and re-mounted it and now
> everything's fine.
> 
> I'm not sure where it got 8:17, though. I guess mtab was stale?

Yah, that would explain it. The values are pulled from mtab via
getmntent()

> 
> Well, I learned a thing about test-qga today: If your local system is
> weird it will explode and get angry at you for your trouble.

Yes, one of it's less-appreciated features :) This specific situation
should be accounted for though. I'll send a patch.

> 
> --js
>