The first revision to try is r264334, which is where I reverted a bad debug info change.

Roll CL: https://codereview.chromium.org/1827363002
This was created with krasin's new script:
$ tools/clang/scripts/upload_revision.py 264334

The pdata thing is figured out, yes? That's the debug info revert?

> The pdata thing is figured out, yes? That's the debug info revert?

No, the debug info revert was for the iOS build.

We haven't been able to reproduce the .pdata problem yet. Our thinking was that unbreaking the asan build is worth risk of potentially breaking the 64-bit win-clang dbg build.

Hm that's the config our fledgling try slaves have... But ok, I wanted to move those to a different config anyhow. Ignore me, sorry :-)

(64bit dbg is also the default gn config for devs I think)

Maybe it's just that bot having problems? I've comments on  crbug.com/595702 , it built fine after I nuked the out/ dir.


The trybot failed to package Clang for Windows due to a test failure:
"Reading PDBs is not supported on this platform."

That fails locally for me too. Also, this UBSan suppression test fails:
Command 7 Stderr:
D:\src\chromium\src\third_party\llvm\projects\compiler-rt\test\ubsan\TestCases\I
nteger\suppressions.cpp:29:44: runtime error: unsigned integer overflow: 1000000
0000000000000 + 9000000000000000000 cannot be represented in type 'unsigned long
 long'

Presumably because it also relies on DIA for symbolization?                                                                              

Hm, the 2015 depot tools bundle has dia SDK and sets VSINSTALLDIR correctly, I think. That problem has its own bug as blocker for krasin's roll but bug iirc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/9be8ca7b77549c46f9a51e991d25e1a509664fd5

commit 9be8ca7b77549c46f9a51e991d25e1a509664fd5
Author: Reid Kleckner <rnk@google.com>
Date: Fri Mar 25 16:53:10 2016

Add -fms-compatibility-version=19 to the clang-cl self-host

Otherwise, we get errors about char16_t and char32_t during the stage 2
build.

Also, fix a bug where cflags, cxxflags, and ldflags all pointed to the
same list.
R=hans@chromium.org
BUG= 597854 

Review URL: https://codereview.chromium.org/1832093002 .

Cr-Commit-Position: refs/heads/master@{#383288}

[modify] https://crrev.com/9be8ca7b77549c46f9a51e991d25e1a509664fd5/tools/clang/scripts/update.py

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/260528431247b958483c9f7f2e1c6525aba077d7

commit 260528431247b958483c9f7f2e1c6525aba077d7
Author: rnk <rnk@chromium.org>
Date: Fri Mar 25 18:10:06 2016

Statically link the CRT so that clang-cl can run in wine on goma

Now we have these deps:
$ dumpbin /dependents third_party\llvm-bootstrap\bin\clang-cl.exe
...
    SHELL32.dll
    KERNEL32.dll
    ADVAPI32.dll

R=hans@chromium.org
BUG= 597854 
NOTRY=True

Review URL: https://codereview.chromium.org/1832113002

Cr-Commit-Position: refs/heads/master@{#383311}

[modify] https://crrev.com/260528431247b958483c9f7f2e1c6525aba077d7/tools/clang/scripts/update.py

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/22c8852e8f91c670cb7805229676b57810725d2a

commit 22c8852e8f91c670cb7805229676b57810725d2a
Author: hans <hans@chromium.org>
Date: Mon Mar 28 19:24:22 2016

Roll clang 263324:264334

BUG= 597854 

Review URL: https://codereview.chromium.org/1827363002

Cr-Commit-Position: refs/heads/master@{#383536}

[modify] https://crrev.com/22c8852e8f91c670cb7805229676b57810725d2a/tools/clang/scripts/update.py

This is done. It caused  bug 598849 , but that doesn't seem revert-worthy.

We also have an ASan issue in  bug 598943 . Fixing that will require rebuiling and rolling forward. Rolling back won't help, so I'm declaring this fixed.