Re: [PATCH v4 3/9] tests/functional: Provide GDB to the functional tests

["Daniel P. Berrangé" <berrange@redhat.com>]

On Mon, Sep 29, 2025 at 08:34:03AM +0200, Thomas Huth wrote:
> On 26/09/2025 20.15, Gustavo Romero wrote:
> > Hi Thomas,
> > 
> > On 9/26/25 15:08, Gustavo Romero wrote:
> > > Hi Thomas,
> > > 
> > > On 9/26/25 07:03, Thomas Huth wrote:
> > > > On 26/09/2025 07.15, Gustavo Romero wrote:
> > > > > The probe of GDB is done in 'configure' and the full path is passed to
> > > > > meson.build via the -Dgdb=option.
> > > > > 
> > > > > Because a single functional test can cover different arches, such as
> > > > > aarch64, ppc64, and x86_64, only a GDB that supports all the arches in
> > > > > the target list is passed to Meson for use in the functional tests. To
> > > > > handle this check, a new shell function, is_target_arch_in_arch_list, is
> > > > > introduced in 'configure'.
> > > > > 
> > > > > Meson then can pass the location of GDB to the test via an environment
> > > > > variable: QEMU_TEST_GDB.
> > > > > 
> > > > > Signed-off-by: Gustavo Romero <gustavo.romero@linaro.org>
> > > > > Signed-off-by: Thomas Huth <thuth@redhat.com>
> > > > > ---
> > > > >   configure                     | 21 +++++++++++++++++++++
> > > > >   meson_options.txt             |  2 ++
> > > > >   scripts/meson-buildoptions.sh |  2 ++
> > > > >   tests/functional/meson.build  |  6 ++++++
> > > > >   4 files changed, 31 insertions(+)
> > > > > 
> > > > > diff --git a/configure b/configure
> > > > > index 0f7eb95586..20e05d233f 100755
> > > > > --- a/configure
> > > > > +++ b/configure
> > > > > @@ -1142,12 +1142,31 @@ fi
> > > > >   #########################################
> > > > >   # gdb test
> > > > > +# Check if all target arches are in a provided list of arches.
> > > > > +is_target_arch_in_arch_list() {
> > > > > +    arch_list=$1
> > > > > +    for target in $target_list; do
> > > > > +        arch=${target%%-*}
> > > > > +        if test "${arch_list#*$arch}" = "$arch_list"; then
> > > > > +            # Target arch not in arch list
> > > > > +            return 1
> > > > > +        fi
> > > > > +    done
> > > > > +    return 0
> > > > > +}
> > > > > +
> > > > >   if test -n "$gdb_bin"; then
> > > > >       gdb_version_string=$($gdb_bin --version | head -n 1)
> > > > >       # Extract last field in the version string
> > > > >       gdb_version=${gdb_version_string##* }
> > > > >       if version_ge $gdb_version 9.1; then
> > > > >           gdb_arches=$($python
> > > > > "$source_path/scripts/probe-gdb- support.py" $gdb_bin)
> > > > > +
> > > > > +    if is_target_arch_in_arch_list "$gdb_arches"; then
> > > > 
> > > > No TABs, please!
> > > > 
> > > > > +            gdb_multiarch="yes"
> > > > > +        else
> > > > > +            gdb_multiarch=""
> > > > > +    fi
> > > > 
> > > > This unfortunately does not work with the GDB from Fedora - it
> > > > only supports "arch64_be arm riscv64 riscv32 ppc i386 s390x
> > > > ppc64 aarch64 ppc64le x86_64", but if you configured a target
> > > > like "alpha-softmmu", this breaks.
> > > 
> > > argh! ok
> > > 
> > > 
> > > > (BTW, does the gdb-multiarch from Debian/Ubuntu really also
> > > > support exotic QEMU targets like tricore?)
> > > 
> > > No, I've checked GDB upstream and I can't see any trace of tricore.
> > > And I just saw that Alex left a comment in scripts/probe-gdb-support.py
> > > saying "# no tricore in upstream gdb", so nope, it seems that it still holds.
> > > 
> > > 
> > > > I think it would be better to drop this hunk, and rather check
> > > > in the spot where we use GDB if the required target is really
> > > > there (i.e. in the functional test that uses it).
> > > 
> > > OK. I'm also not a big fan of doing it in bash. How do you suggest
> > > to do it? Directly in the code, via a skipIf decorator, or something else?
> > 
> > $gdb_arches, obtained using scripts/probe-gdb-support.py in configure,
> > could be passed to meson and meson sets it in the test env, as we're
> > doing for $gdb_bin and QEMU_TEST_GDB env var. wdyt?
> 
> That might be a possibility, though it's getting a little bit clunky if you
> want to run the test manually, without the meson test runner.
> 
> Maybe it would be nicer to just start gdb in the test and catch the error
> (and skip the test in that case) in the python test code if it fails to set
> the target architecture there?

If you run 'set architecture' with no arguments, it'll return a long list
of supported targets we can match on.


With 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 :|