On Wed, 19 Nov 2025 at 11:14:32 +0100, Santiago Vila wrote:
I confirm that this is fixed in the above version.
And yes, it was not obvious what actually failed. However, I can see
what failed when I look at a build log made on a machine with 1 CPU,
the relevant part of the build log seems to be like this:
make -f CMakeFiles/wrappers.dir/build.make CMakeFiles/wrappers.dir/build
make[3]: Entering directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
[ 40%] Generating wrapper code for: GObject-2.0;Gio-2.0;Gst-1.0;Gtk-3.0
cd /<<PKGBUILDDIR>> && /<<PKGBUILDDIR>>/obj-x86_64-linux-gnu/cppgir --class
--class-full --ignore data/cppgir.ignore:d
ata/cppgir_unix.ignore --output /<<PKGBUILDDIR>>/debian/tmp GObject-2.0 Gio-2.0
Gst-1.0 Gtk-3.0
ERROR could not find GIR for Gst-1.0
I wonder why this wasn't visible in the log you initially linked, which
was presumably from a machine with more than one CPU? I would expect
that a parallel build could interleave warning/error output from more
than one process, but I would also have expected the error to turn up
*somewhere* in the log.
"ERROR could not find GIR for Gst-1.0" looks like a symptom of #1060906
(older cppgir didn't look at the Debian-specific multiarch GIR XML
search paths), combined with libgstreamer1.0-dev (>= 1.26.6-2)
installing Gst-1.0.gir into the multiarch GIR XML directory so that it
can be Multi-Arch: same again. So it does make sense that this newer
version, which searches all the right directories, fixes the FTBFS.
smcv
