Summary: | The generated commit message have a directory label at the first line rather than the bug's title if ChangeLogs have different bug titles | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Fujii Hironori <Hironori.Fujii> | ||||||||
Component: | Tools / Tests | Assignee: | Fujii Hironori <Hironori.Fujii> | ||||||||
Status: | RESOLVED WONTFIX | ||||||||||
Severity: | Normal | CC: | ap, darin, ddkilzer, ews-watchlist, glenn, jbedard, mcatanzaro, simon.fraser, webkit-bug-importer | ||||||||
Priority: | P2 | Keywords: | InRadar | ||||||||
Version: | WebKit Nightly Build | ||||||||||
Hardware: | Unspecified | ||||||||||
OS: | Unspecified | ||||||||||
See Also: | https://bugs.webkit.org/show_bug.cgi?id=220950 | ||||||||||
Attachments: |
|
Description
Fujii Hironori
2021-01-21 13:30:25 PST
This is not Subversion friendly either. This has to be a regression I presume? Created attachment 418245 [details]
Patch
(In reply to Alexey Proskuryakov from comment #1) > This has to be a regression I presume? I think not. Rather, a case we did not optimize for. Created attachment 418248 [details]
Patch
Comment on attachment 418248 [details]
Patch
Really need regression tests for this script. review+ on "faith" since I didn’t really study it carefully enough to spot mistakes.
Comment on attachment 418248 [details] Patch Clearing flags on attachment: 418248 Committed r271805: <https://trac.webkit.org/changeset/271805> All reviewed patches have been landed. Closing bug. A Git pull request has a commit message. I hope WebKit will deprecate ChangeLog after the Git migration. (In reply to Fujii Hironori from comment #9) > A Git pull request has a commit message. > I hope WebKit will deprecate ChangeLog after the Git migration. The social contract that goes with our decision of including change logs vs. relying only on commit messages (and note that Subversion has commit messages just like Git does and always has, as did CVS before it) goes far beyond the considerations in making this script work well. Reverted r271805 for reason: Appears to have broken generation of changed files/functions in commit logs Committed r271865: <https://trac.webkit.org/changeset/271865> (In reply to Ryan Haddad from comment #11) > Reverted r271805 for reason: > > Appears to have broken generation of changed files/functions in commit logs It also caused some webkitpy test failures, documented in https://bugs.webkit.org/show_bug.cgi?id=220950 Besides the test failures, this caused problems for some programmers at Apple who found the script wasn’t including files at all any more in their configuration. There’s a chance that’s specific to Apple’s internal configuration. Created attachment 418363 [details]
Patch
(In reply to Darin Adler from comment #13) > Besides the test failures, this caused problems for some programmers at > Apple who found the script wasn’t including files at all any more in their > configuration. There’s a chance that’s specific to Apple’s internal > configuration. I fixed the test failure. And webkitpy EWS passed. However, I don't know about the Apple internal issue. It seems that I should give up. I hope someone in Apple would take over this task. (In reply to Fujii Hironori from comment #15) > (In reply to Darin Adler from comment #13) > > Besides the test failures, this caused problems for some programmers at > > Apple who found the script wasn’t including files at all any more in their > > configuration. There’s a chance that’s specific to Apple’s internal > > configuration. > > I fixed the test failure. And webkitpy EWS passed. > However, I don't know about the Apple internal issue. It seems that I should > give up. I hope someone in Apple would take over this task. It seems unlikely that the Apple Internal issue is actually special, I suspect that the webkitpy test failure caught the mistake, we just didn't have EWS configured properly to catch it (Aakash Jain fixed that yesterday). Will double-check today what the specifics of the Apple Internal issue was. ChangeLog has been deprecated. Closed as WONTFIX. |