Re: [PATCH 14/14] iotests: make img_info_log() call qemu_img_log()

[John Snow]

On Thu, Mar 17, 2022 at 1:00 PM John Snow <jsnow@redhat.com> wrote:
>
> On Thu, Mar 17, 2022 at 11:39 AM Hanna Reitz <hreitz@redhat.com> wrote:
> >
> > On 09.03.22 04:54, John Snow wrote:
> > > Add configurable filters to qemu_img_log(), and re-write img_info_log()
> > > to call into qemu_img_log() with a custom filter instead.
> > >
> > > After this patch, every last call to qemu_img() is now guaranteed to
> > > either have its return code checked for zero, OR have its output
> > > actually visibly logged somewhere.
> > >
> > > Signed-off-by: John Snow <jsnow@redhat.com>
> > > ---
> > >   tests/qemu-iotests/iotests.py | 13 +++++++++----
> > >   1 file changed, 9 insertions(+), 4 deletions(-)
> >
> >  From my POV, this is a regression because before this patch (not this
> > series, though, admittedly), `img_info_log()` would throw an exception
> > on error, and with patch 12 being as it is, it will revert to its
> > pre-series behavior of not throwing an exception.  I prefer exceptions

Oh, actually... patch #12 does this:

-    output = qemu_img_pipe(*args)
+    output = qemu_img(*args, check=False).stdout

so I never actually toggled error checking on for this function at
all. This isn't a regression.

At a glance, img_info_log() calls fail as a matter of course in 242
and 266 and ... hm, I can't quite test 207, it doesn't work for me,
even before this series.

I didn't test *all* qemu_img calls yet either, but ... I'm going to
gently suggest that "converting logged calls to qemu_img() to be
checked calls" is "for another series" material.

--js

> > to failed reference output diffs, because an exception tells me which
> > call failed.
>
> Hm, yeah. I just need to figure out if *all* of the qemu_img_log()
> calls are safe to enforce the return code of zero on... or how many
> need work if I change the default behavior. Let me see what I can do.
>
> I suppose it's maybe a bit late to try and squeak any of this in for
> freeze, so I can roll everything back up into one big series again and
> send a new revision.