Closed Bug 1802359 Opened 6 months ago Closed 4 months ago

mozregression GUI doesn't work on macOS 13.0.1


(Testing :: mozregression, defect, P2)



(Not tracked)



(Reporter: soeren.hentzschel, Assigned: zeid)




(1 file)

Attached image mozregression.png

Even after bug 1781111 and with mozregressiom 5.2.0 the mozregression GUI doesn't work on macOS 13.0.1. At least I found a way to start the builds, but then it crashes after the step to decide if the build was good or bad.

The attached image shows the different states. The language is German, sorry for that.

  • When mozregression wants to start the first build, a dialog appears, saying that Firefox Nightly is damaged and can't be opened. There is no button to open the application (1).
  • Then an entry in the security section of the macOS settings appears. There is an button to open Firefox Nightly (2).
  • After pressing that button a new dialog appears. This time there is a button to open the build (3).
  • The build starts.
  • Press either the "Good" or the "Bad" button in the mozregression windows, then a crash dialog appears. It doesn't matter wich button you press, you won't get the next build (4).

platform: macOS-10.16-x86_64-i386-64bit
python: 3.9.1 FROZEN (64bit)
mozregression: 5.2.0
message: PermissionError: [Errno 1] Operation not permitted: '/var/folders/v0/89gzxmpj13s7q05_r16p6_c00000gn/T/tmptwkwosvb/Firefox'
traceback: File "mozregui/", line 264, in set_verdict
File "mozregui/", line 82, in finish
File "/Applications/mozregression", line 323, in cleanup
File "mozfile/", line 231, in remove
File "mozfile/", line 214, in _update_permissions
File "mozfile/", line 191, in _call_with_windows_retry
File "mozfile/", line 149, in _call_windows_retry

With the mozregression command line tool there are no problems.

Severity: -- → S3
Priority: -- → P2
Assignee: nobody → zeid

I believe this is happening because mozregression-gui is running under Rosetta (we don't have arm64 or universal2 builds yet), and under Rosetta, mozregression thinks it's running on macOS 10.6, which makes it ignore new codesign requirements. Short-term solution is to force code re-signing if we're running in Rosetta. Long term solution is to have universal2 builds and/or arm64 builds.

See Also: → 1781111
Depends on: 1810220
Depends on: 1810534

Fixed in 5.3.0.rc1. Will be released as 5.3.0 soon.

Closed: 4 months ago
Resolution: --- → FIXED

Thank you! I can confirm that it's fixed (tested on MacBook Pro M1 Pro, macOS 13.1).

You need to log in before you can comment on or make changes to this bug.