On Mon, Jan 02 2023, Ihor Radchenko wrote: > Ihor Radchenko writes: > >> Ihor Radchenko writes: >> >>>> There is a race condition between writing the contents of the graphics >>>> file to disk and emacs checking the file size. My guess is that this is >>>> causing the problem (and that the same failure applies for emacs-2{6,7}, >>>> since only the emacs-28 reports the exact test failure). >>> >>> Maybe we can just add several `sleep-for' calls to the test? >> >> I just did this. Let's see if CI errs again. >> https://git.savannah.gnu.org/cgit/emacs/org-mode.git/commit/?id=e5c45358a > > Apparently, `sleep-for' 1 second was not enough, and I decided to remove > checking file size completely. Hello Ihor, Is there an environment variable that could be used to determine is the tests are being run on sourcehut? This would let us cut out that test on sourcehut, while still keeping it elsewhere. > > Upon doing this, another failure popped up. This time, it looks like an > actual Elisp issue: > > https://builds.sr.ht/~bzg/job/914954 > 2 unexpected results: > FAILED ob-octave/graphics-file ((should-not (get-buffer "*Org-Babel > Error Output*")) :form (get-buffer "*Org-Babel Error Output*") :value > #) > FAILED ob-octave/graphics-file-space ((should-not (get-buffer > "*Org-Babel Error Output*")) :form (get-buffer "*Org-Babel Error > Output*") :value #) > > As you can see *Org-Babel Error Output* buffer does not exist when > running the test. > > Leo, could you please take a look? An earlier test is creating that *Org Babel Error Output* buffer. That is killed on the first test, before the test is actually run. But GET-BUFFER behaves in an undocumented way: it returns a non-nil value, #. To remedy that, I have wrapped the calls in BUFFER-LIVE-P. See the attached patch. Leo