GNOME Bugzilla – Bug 390673
videocrop check fails in iurt build bot
Last modified: 2006-12-30 12:58:22 UTC
This is on Mandriva Cooker i586. In an interactive login the build and checks of gst-plugins-bad 0.10.4 all succeed, but in the iurt build bot one check fails: elements/videocrop.c:179:F:general:test_crop_to_1x1: Assertion 'ctx->src != NULL' failed elements/videocrop.c:179:F:general:test_caps_transform: Assertion 'ctx->src != NULL' failed elements/videocrop.c:179:F:general:test_passthrough: Assertion 'ctx->src != NULL' failed elements/videocrop.c:84:F:general:test_unit_sizes: Assertion 'csp != NULL' failed elements/videocrop.c:179:F:general:test_cropping: Assertion 'ctx->src != NULL' failed FAIL: elements/videocrop I'll attach the build logs.
Created attachment 79029 [details] failed check from the iurt build bot log
Created attachment 79030 [details] successful build and checks from an interactive session
This means it can't find and/or create a videotestsrc element and an ffmpegcolorspace element, both of which are in gst-plugins-base. No idea why that would be. Do the build dependencies include all -base packages including those containing the plugins? Can you make the buildbot build the package with a GST_DEBUG=*:2 or GST_DEBUG=*:3 environment somehow? (plus GST_DEBUG_NO_COLOR=1 preferably)
Good catch. The build system will only install packages that are in BuildRequires, gst-plugins-base was not (only the -devel part of it). Maybe the configure script should check for the videotestsrc element, what do you think?
> The build system will only install packages that are in > BuildRequires, gst-plugins-base was not (only the -devel part of it). > Maybe the configure script should check for the videotestsrc element, > what do you think? Not really sure, since a) many GStreamer developers feel checking for elements in configure scripts is generally not useful, because it has to be done again at runtime anyway b) most people never run 'make check' c) if we do add the check, other people that never actually run 'make check' as part of their build/buildbot setup will complain about the unnecessary build dependency ;) d) arguably, we only really ship source tarballs as a whole, so by implication the check for the -base headers should be sufficient (and everything else would be either a packaging issue or someone mixing sources and packages who then should better know what she's doing). I think it's best to just make the failure cases a bit more verbose so that it's more obvious what's missing: 2006-12-30 Tim-Philipp Müller <tim at centricular dot net> * tests/check/elements/videocrop.c: (GST_START_TEST), (videocrop_test_cropping_init_context): When we can't create an element needed for the test, print a message detailing which element it actually is that's missing (#390673).