Buildbot failures

A recent fedora-gcc build (http://buildbot.octave.org:8010/#/builders/11/builds/2665) failed with the following error:

error: print: figure must be visible or qt toolkit must be used with __gl_window__ property 'on' or QT_OFFSCREEN feature available
error: called from
    __opengl_print__ at line 209 column 7
    print at line 771 column 16
    plotimages at line 113 column 7
make[2]: *** [Makefile:31919: doc/interpreter/extended.pdf] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory '/home/buildbotu/fc25-x86_64/gcc-fedora/build'
make[1]: *** [Makefile:28026: all-recursive] Error 1
make[1]: Leaving directory '/home/buildbotu/fc25-x86_64/gcc-fedora/build'
make: *** [Makefile:11464: all] Error 2

I’m a bit confused about why that would happen. The build log begins with

xvfb-run -a -s '-screen 0 640x480x24' make V=1 -j4
 in dir /home/buildbotu/fc25-x86_64/gcc-fedora/build (timeout 1200 secs)
 watching logfiles {}
 argv: [b'xvfb-run', b'-a', b'-s', b'-screen 0 640x480x24', b'make', b'V=1', b'-j4']
...

so shouldn’t all of the subsequent commands executed in this build step be using the xvfb framebuffer to provide a virtual graphical display?

IIRC, I saw similar issues on the GitHub runners. The “solution” was to build with -j1.

I’m not sure what is happening. But I was under the impression that xvfb sometimes failed to provide a (virtual) display when loaded with lots of simultaneous commands…

Edit: That was here: CI with GitHub hosted runners - #7 by mmuetzel

Oh, that could certainly explain why this failure is random.

I think all those failures happened under Wayland. I switched to xorg for now.

p.s. I checked ans some of those were under X11 (you can check for XDG_SESSION_TYPE in the logs).

OK. I also pushed the following change to provide some more info if/when errors like this happen again:

http://hg.savannah.gnu.org/hgweb/octave/rev/a87e5f9d5446