Closed Bug 1304042 Opened 3 years ago Closed 2 years ago
buildsymbols should fail if running dsymutil fails
59 bytes, text/x-review-board-request
I'm not sure why, but currently the buildsymbols step doesn't fail on mac if running dsymutil fails. This means that bug 1301751 slipped in and we didn't notice because the builds didn't break. :-(
It's fairly likely that this try push will have broken builds because it's exposing existing brokenness. :-(
...or it could fail because I typoed the name of the "success" variable. Also apparently concurrent.futures doesn't handle sending a `CalledProcessException` back from a job, so it still failed.
As usual, it's never as easy as it seems. I fixed the things in comment 6, but I also had to fix the invocation of symbolstore.py in Makefile.in, because it was being invoked in a shell pipeline, so its exit status was being ignored. :-( I tested locally with a dsymutil binary in my path that fails and `mach buildsymbols` correctly exits with an error in that case.
(In reply to Ted Mielczarek [:ted.mielczarek] from comment #7) > https://treeherder.mozilla.org/#/jobs?repo=try&revision=491d798446c8 The opt build here is correctly red. llvm-dsymutil crashed and buildsymbols failed because of that.
Comment on attachment 8792901 [details] bug 1304042 - buildsymbols should fail if running dsymutil fails. https://reviewboard.mozilla.org/r/79766/#review79204 ::: Makefile.in:316 (Diff revision 3) > $(PYTHON) $(topsrcdir)/toolkit/crashreporter/tools/symbolstore.py \ > $(MAKE_SYM_STORE_ARGS) \ > $(foreach dir,$(SYM_STORE_SOURCE_DIRS),-s $(dir)) \ > $(DUMP_SYMS_BIN) \ > $(DIST)/crashreporter-symbols \ > - $(MAKE_SYM_STORE_PATH) | grep -iv test > \ > + $(MAKE_SYM_STORE_PATH) > \ Nit: trailing whitespace
Attachment #8792901 - Flags: review?(gps) → review+
https://hg.mozilla.org/integration/mozilla-inbound/rev/f82827622145bcd160f63ec48e9606498beefc39 bug 1304042 - buildsymbols should fail if running dsymutil fails. r=gps
Backed this out for making failures in buildsymbols like https://treeherder.mozilla.org/logviewer.html#?job_id=36546605&repo=mozilla-inbound cause the build to fail. https://hg.mozilla.org/integration/mozilla-inbound/rev/689f6fee1846
So this got backed out for exposing two (existing) failures: 1) The tier-2 taskcluster OS X opt builds are failing to run llvm-dsymutil on the gtest XUL library. This is unfortunate and I don't know of a way to make this work, but hopefully bug 1304815 will fix it when it lands. 2) There were some odd failures in Windows builds, like: https://treeherder.mozilla.org/logviewer.html#?job_id=36549310&repo=mozilla-inbound#L19896 12:26:41 INFO - Unexpected error: [Error 2] The system cannot find the file specified: 'dist\\crashreporter-symbols\\shlibsign.pdb\\5BCF415274384AF4A8FBD848F550CC3A1\\shlibsign.pdb' I don't know quite what's happening there, but I'll have to sort that out before relanding this.
Depends on: 1304815
I'll rebase the patch here.
Assignee: ted → cmanchester
(In reply to Chris Manchester (:chmanchester) from comment #16) > I'll rebase the patch here. This is failing where it should and nowhere it shouldn't on try.
Comment on attachment 8886330 [details] Bug 1304042 - Fail the build if running dsymutil fails. https://reviewboard.mozilla.org/r/157070/#review162216 LGTM - I assume we can't land this until after bug 1380381 is fixed?
Attachment #8886330 - Flags: review?(mshal) → review+
(In reply to Michael Shal [:mshal] from comment #19) > Comment on attachment 8886330 [details] > Bug 1304042 - Fail the build if running dsymutil fails. > > https://reviewboard.mozilla.org/r/157070/#review162216 > > LGTM - I assume we can't land this until after bug 1380381 is fixed? Right. I guess the trees are closed for it anyway :/
Depends on: 1380381
Pushed by email@example.com: https://hg.mozilla.org/integration/autoland/rev/4c8ee44afffb Fail the build if running dsymutil fails. r=mshal
Attachment #8792901 - Attachment is obsolete: true
You need to log in before you can comment on or make changes to this bug.