Closed oliv3r closed 8 months ago
Here'st he interposer log from the pipeline, probably best to attach it to the issue so it don't get lost.
with the most interesting bits here:
PANGO:tmp2gwix2eu.png
PANGO:#
KiAuto:match
KiAuto:Waiting for `['GLX:Swap']` starts=True times=1
IO:open:/tmp/.icons/default/cursors/left_ptr
IO:fopen:r
IO:open:/tmp/.icons/default/index.theme
IO:fopen:r
IO:open:/usr/share/icons/default/cursors/left_ptr
IO:fopen:r
IO:open:/usr/share/icons/default/index.theme
IO:fopen:r
IO:open:/usr/share/pixmaps/default/cursors/left_ptr
IO:fopen:r
IO:open:/usr/share/pixmaps/default/index.theme
IO:fopen:r
KiAuto:Waiting for `['GLX:Swap']` starts=True times=1
KiAuto:Waiting for `['GTK:Window Destroy:3D Image File Name']` starts=False times=1
PANGO:Appearance
PANGO:Selection Filter
PANGO:Appearance
PANGO:Selection Filter
* wrapping file chooser get filename
GTK:Filename:/builds/olliver/lapod/breakout_hdmi/outputs/render3d/breakout_hdmi-3D_bottom.png
GTK:Filename:**Changed from /builds/olliver/lapod/breakout_hdmi/#
GTK:Main:Out
GTK:Window Set Modal:3D Image File Name 0
GTK:Filename:/builds/olliver/lapod/breakout_hdmi/outputs/render3d/breakout_hdmi-3D_bottom.png
GTK:Filename:**Changed from /builds/olliver/lapod/breakout_hdmi/#
This is the same as #544
When running the following yaml file in the latest kibot 7 full container in a CI, kicad gets confused and fails to save the file (see video in artifacts). Because of that (filename is a # only) kibot keeps waiting for the timeout, and then kills stuff.
Instead, kibot could check the intended filename, and see if it is valid at all.
I've noticed this issue in more flows, so I wonder, would it not be wiser, to save the file as a temporary, mktmpdir in python, check if all went well, store file with a random name there, and then when finished mv the tmp file to the intended destination? This would remove quite a few of these errors, in that the GUI kicad will have a super reliable way to generate the needed file ...
See https://gitlab.com/olliver/lapod/-/jobs/5836088538 and/or https://gitlab.com/olliver/lapod/-/pipelines/1122395553
What's interesting from the logs, is that kibot actually tries to name the file '#'
but why? I think that confuses kicad, though the dialog does exit normally. The vrml issue also does the same, but doesn't print that in its logs ... later though, the timeout kill is of course on the actual filename, which wasn't used of course ...