[Libosinfo] [PATCH] rhel: Fix isodata for RHEL 7.5 on ppc64le

[Daniel P. Berrangé]

On Fri, May 04, 2018 at 04:00:25PM +0200, Andrea Bolognani wrote:
> On Fri, 2018-05-04 at 09:42 +0200, Fabiano Fidêncio wrote:
> > > Now /isodetect/rhel passes, and I get
> > > 
> > >   /isodetect/opensuse:
> > >   ** (/home/test/libosinfo/tests/.libs/lt-test-isodetect:24007): ERROR **: ISO openSUSE-Leap-42.3-NET-x86_64.iso.txt was not matched by OS opensuse42.3
> > >   Trace/breakpoint trap (core dumped)
> > > 
> > > instead. Progress? :)
> > 
> > Sorry for the long email.
> > 
> > Here's the output of how I check that our tests are passing:
> > 
> > [fidencio at machado osinfo-db]$ make
> >   I18N      data/datamap/microsoft.com/win-7-l10n-language.xml
> [...]
> >   GEN       data/schema/osinfo.rng
> >   EXP       osinfo-db-20180504.tar.xz
> >   GEN       osinfo-db.spec
> >   GEN       mingw-osinfo-db.spec
> > [fidencio at machado osinfo-db]$ export OSINFO_SYSTEM_DIR=$PWD/data
> > [fidencio at machado osinfo-db]$ cd ../libosinfo/
> > [fidencio at machado libosinfo]$ make check
> [...]
> > make  check-TESTS
> > make[2]: Entering directory '/home/fidencio/src/upstream/libosinfo/tests'
> > make[3]: Entering directory '/home/fidencio/src/upstream/libosinfo/tests'
> > PASS: test-entity
> [...]
> > PASS: test-isodetect
> > PASS: test-install-script
> > ============================================================================
> > Testsuite summary for libosinfo 1.2.0
> > ============================================================================
> > # TOTAL: 15
> > # PASS:  15
> > # SKIP:  0
> > # XFAIL: 0
> > # FAIL:  0
> > # XPASS: 0
> > # ERROR: 0
> > ============================================================================
> 
> Your step are not identical, but equivalent, to the ones I use.
> Following your instructions to the letter didn't help.
> 
> Weirdly, I'm only seeing this failure on Fedora 27: I tried with
> Debian 9 and Fedora 28 and it worked fine both times, so there's
> probably something wrong with my guest rather than with libosinfo :)

Last time I investigated this kind of wierd problem, I found that it was
caused by non-deterministic file ordering from readdir(), combined with
multiple matches by different osinfo files. That should have been fixed
by:

commit aa06a8b2905f7ee488b1215e35e486cd36c2df26
Author: Daniel P. Berrangé <berrange at redhat.com>
Date:   Fri Mar 16 13:18:22 2018 +0000

    loader: process files in alphabetical order
    
    Currently when loading DB files we process them in whatever order
    readdir() returns them in, which is often inode order. This makes the
    order of loading files non-deterministic across installs, and if there
    are ambiguities in the data, we can in turn get different results.
    eg ISO images match different OS entries depending on load order.
    
    Alphabetically sorting the files doesn't remove any ambiguity that
    may exist, but at least gives us consistent results on every host.
    
    Signed-off-by: Daniel P. Berrangé <berrange at redhat.com>
    Reviewed-by: Fabiano Fidêncio <fabiano at fidencio.org>



Is it possible that your $HOME/.config/osinfo directory has some
stale content perhaps.


Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|