-
Notifications
You must be signed in to change notification settings - Fork 1.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
3.6.0 (darkroom) unstable on Gentoo Linux #9815
Comments
I did some more testing and the mask manager sort-of-works if one adds another shape, a circle for example, and then select the drawn mask (path). That is, it crashes less often if the added shape is not last in the list of shapes it seems. And it also helps to go back to the lightable and then back to darkroom so that the changes are written to disk before it crashes... |
Hi, Your report has not enough data and can't replicate your problem following outlined steps (and I do have gentoo). Can you attach crash logs, xmp files and your system description (ram, opencl etc)? |
Here is my system info:
Graphics; Nvidia 3080 Ti
This is the most unstable version I have tested so far and I get frequent crashes by just pressing the d key in light table. I will try to build 3.6.0 from sources with debug symbols to see if I can get better traces of crashes. |
I have now done some initial tests and it looks like the "crash when pressing d in lighttable" is related to C-compiler flags. I used this config:
which resulted in
after pressing d on a few thumb nails in lighttable. I have
If building with just -O3 (no -march=native) in the C-flags (CXX works with -march=native):
then I cannot generate this crash. The mask manager also seems more stable but I need to test this more. For reference, the Gentoo ebuild uses this config
|
And this is why we need backtraces:
because now we can correlate it with #9340 and #9002 and #8695 (comment) Most important thing here: #9002 (comment) Current master does not exhibit crashes and there are no crashes when compilled with GCC10. The problem is somewhere in GCC probably, I opened a bug report about it here: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101262 If you could create reproducer it would be awesome. |
For reference, since I have gcc-10.3 and clang-12 on the machine too: First try gcc-10.3.0
This builds and runs fine. Got no crashes as with gcc-11 and -march=native. Next try clang-12
Some LDFLAGS fixes this
This also seems to run fine. |
Sooo... It looks like your report is regarding issue that is fixed in newest 11 series GCC, not present before or after 11.x and not present in LLVM. I'd say this is fixed and/or workaround is easily "present" |
@johnny-bit Do you have a link to where this fix for gcc 11.2.0 is discussed? Just curious since I have issues with gcc-10.3.0 too (internal compiler error when building LibreOffice) |
I've opened 2 bugs against gcc. Reproducers for bugs are very important :) |
I get random crashes when editing Nikon .NEF files in 3.6.0 I started clean by deleting:
.config/darktable/data.db
.config/darktable/library.db
I just get random segfaults when pressing "d" in lightable. Restarting darktable and opening another file first sometimes help but not always.
Also, applying a mask to, at least the exposure module, always results in:
darktable: /var/tmp/portage/media-gfx/darktable-3.6.0/work/darktable-3.6.0/src/develop/pixelpipe_hb.c:1470: dt_dev_pixelpipe_process_rec: Assertion `tiling.factor > 0.0f' failed.
Aborted
That is, doing:
I had similar problems with masks in 3.4.1 when the mask extended outside the image but on 3.6.0 it always aborts darktable.
The text was updated successfully, but these errors were encountered: