-
Notifications
You must be signed in to change notification settings - Fork 2.5k
KDE-Thumbnailers: Cleanup $WORK data after build #13944
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
KDE-Thumbnailers: Cleanup $WORK data after build #13944
Conversation
amazingakai has previously contributed to projects/kde-thumbnailers. The previous PR was #13490 |
Is this really needed? Seems like it should be done by default? |
hmm, I'm not entirely sure. Let's see if this fixes it in case. Otherwise, it would have to be an issues at runtime -- do any of the harnesses for some reason dump/log a lot to disk? |
The build is still failing: https://oss-fuzz-build-logs.storage.googleapis.com/log-4433fb3a-d637-44ae-b430-2fcfb05943fc.txt Only the build-check-centipede-* fails with storage error, the previous two (afl and libFuzzer) checks pass successfully. I checked some of the fuzzers, they don't do anything writing to disk except for writing to a temporary file which gets autodeleted when the process exits. Should we maybe try disabling centipede build? |
@DavidKorczynski @tsdgeos ping |
What if the process runs for hours, does that make the temporary files grow and grow? |
No, the temporary files get autodeleted so they shouldn't grow. |
I'll try and debug this tonight |
I still need some more time to fully debug this, but I think I have a solution. In the OSS-Fuzz build logs all the harnesses are failing before the disk exhaustion occurs. However, this is only for the centipede engine (see Locally, however, with the libfuzzer engine all but 1 harness pass:
The 1 breaking harness fails because of a leak. As such, to fix this, I suggest (1) fix/silence the leak; (2) disable centipede for now. I was running some testing with centipede but am atm on a small machine and the linking crashed OOMed my memory and building took a while to compile, so will have to pick this up tomorrow. But I assume that also signals for centipede there's some odd stuff happening which blows up the size of the harnesses. However, I'm quite sure the above two steps (double check other engines fwiw) will make build green. |
The build fails at the
check_build
step due to running out of storage.This change cleans up $WORK after the build to free up space.
https://oss-fuzz-build-logs.storage.googleapis.com/log-150d58fe-14f9-41cc-ac9c-1ae491fe4556.txt