Unassigning from myself, as I'm not able to test any Mac-related changes.

Unfortunately both https://cluster-fuzz.appspot.com/testcase?key=5170337031651328 and https://cluster-fuzz.appspot.com/testcase?key=6350731919753216 are invalid, so I can't tell what's wrong with them.

My wild guess would be that the build system has recently migrated to GN, and something went wrong.

That seems like a very valid guess.

What does "sometimes" mean? I would expect the behavior to be deterministic. Also, which bot is producing the build artifacts being tested? It would be good to know its gn args.

Another guess is that might be a difference between crashes that happen in the Chromium binary and the frameworks (the former is symbolized, the latter is not). At least I've dealt with similar problems in  issue 148383 .

What we need to check is:
 - that we still build with debug info (probably yes, since some of the reports do contain file:line info)
 - that real .dSYM files are produced for every piece of Chromium
 - that llvm-symbolizer is able to find those .dSYM files

It could be that the .dSYM file locations have changed, and we're passing incorrect paths to them from asan_symbolize.py (https://cs.chromium.org/chromium/src/tools/valgrind/asan/asan_symbolize.py?sq=package:chromium&dr=C&l=135)

s/that might be/that there might be/

Here is another report with missing stuff - https://cluster-fuzz.appspot.com/testcase?key=5170535808106496 (and should not get deleted, since associated with bug). https://cluster-fuzz.appspot.com/testcase?key=6173107731824640 is one with line numbers.

Link to builder flags - https://build.chromium.org/p/chromium.lkgr/builders/Mac%20ASAN%20Release/builds/3304/steps/generate_build_files/logs/stdio

Alex - I dont see any .dSYM files in the build archive. Any idea what stuff in gyp was creating that before ?

I think we are missing integration with enable_dsyms for sanitizer builds - https://cs.chromium.org/chromium/src/tools/gn/docs/cookbook.md?rcl=0&l=226
https://cs.chromium.org/chromium/src/build/config/mac/symbols.gni?rcl=0&l=16

Previously we had that =1 in gyp - https://cs.chromium.org/chromium/src/build/common.gypi?rcl=0&l=461.

Dirk, thoughts?

See also  Issue 628052  in case you were using fake dsyms, these are gone and there are now .unstripped files. Real dsyms are probably still around though, so maybe it's just a config problem.

re #10 - yup, looks like you're right, so this should perhaps just change the default for enable_dsyms to (is_chrome_branded || is_asan).

(As an aside, the enable_dsyms default is current (is_official_build), but in theory is_official_build is badly named, since it really reflects the optimization level, not whether or not this is an actual official build, so we should change the default as part of this).

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8b0cda7f8d3f9a4f1d53c78600fb0c62c7eec7f9

commit 8b0cda7f8d3f9a4f1d53c78600fb0c62c7eec7f9
Author: dpranke <dpranke@chromium.org>
Date: Wed Aug 10 06:17:52 2016

Make sure dSYMs are created by default for Mac sanitizer builds.

They should be created by default for sanitizer builds (and were w/
GYP), but that setting got lost in the GN migration.

Also, this CL corrects the default logic for enable_dsyms and
enable_stripping so that we only do this for internal official
builds by default, rather than for all "official" builds (
since is_official_build is really an optimization setting, not
a description of how things should be packaged).

R=rsesek@chromium.org, thakis@chromium.org, inferno@chromium.org
BUG= 635242 

Review-Url: https://codereview.chromium.org/2230723002
Cr-Commit-Position: refs/heads/master@{#410979}

[modify] https://crrev.com/8b0cda7f8d3f9a4f1d53c78600fb0c62c7eec7f9/build/config/mac/symbols.gni

There was one more fix needed. The .dSYMs paths changed and now it does not use .app and .framework at end. Verified and removed in CF source. https://chromereviews.googleplex.com/483457013/

I feel happy that this is fixed, this was really important for FindIt culprit CL finder.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/002fbfa1c30058690e76c010d1b0b8801e8a2d39

commit 002fbfa1c30058690e76c010d1b0b8801e8a2d39
Author: rsesek <rsesek@chromium.org>
Date: Thu Aug 11 12:13:21 2016

Partially revert 8b0cda7f8d3f9a4f1d53c78600fb0c62c7eec7f9.

That CL modified the conditions under which enable_dsyms and enable_stripping
get set, which is not desired. This leaves in place the change for sanitizers.

BUG= 635242 
R=dpranke@chromium.org

Review-Url: https://codereview.chromium.org/2232943003
Cr-Commit-Position: refs/heads/master@{#411306}

[modify] https://crrev.com/002fbfa1c30058690e76c010d1b0b8801e8a2d39/build/config/mac/symbols.gni