I saw this while building LLVM, but http://llvm.org/viewvc/llvm-project?view=revision&revision=273294 made it go away for me.

I think this won't be the last time we emit corrupt codeview, so it might be worth automating the process of deleting corrupt PDBs on the TOT bots. =/

Well, for better or worse this still reproduces with clang:

[163/163] Linking CXX executable bin\clang.exe
FAILED: bin/clang.exe
...
LINK : fatal error LNK1285: corrupt PDB file 'D:\src\llvm\build_debug\bin\clang.pdb'; delete and rebuild

Ok, so I guess this is just a known bug then, and the bot will green up once it's fixed.

Does this block clang rolls? the asan lkgr bots use pinned clang and build with debug info too, right?

> Ok, so I guess this is just a known bug then, and the bot will green up once it's fixed.

The TOT bots don't do a full clobber, right? They just use a unique -D flag to invalidate the ninja cache? If so, the existing corrupt PDB will be there in place until we remove it, so the bot won't turn green by itself. We may want to take the time to really clobber PDBs, because we can expect LLVM will generate corrupt codeview from time to time.

> Does this block clang rolls? the asan lkgr bots use pinned clang and build with debug info too, right?

ASan uses pinned clang, so yes, I would consider this a blocker for a clang roll.

adding blocker then. i wanted to roll today (it's been a month), guess i'll wait a bit.

It's very easy to temporarily make the bot a real clobber bot for a few builds once you think the bug's fixed. (we just have to add 'clobber' to the bot's recipe config)

There was a remaining bug in large record splitting that I fixed in r273477. Can we clobber the asan builder and see if it goes away and stays away?

sure thing: https://codereview.chromium.org/2092553002/

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/tools/build.git/+/73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5

commit 73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5
Author: thakis <thakis@chromium.org>
Date: Wed Jun 22 22:51:55 2016

Make CrWinASan* bots true clobber bots.

They do clang builds with debug info, and incremental links mean that
pdb files might be in confused states while we work on debug info. Make
sure the pdb files are deleted each time.

BUG= 622339 

Review-Url: https://codereview.chromium.org/2092553002

[modify] https://crrev.com/73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5/scripts/slave/recipe_modules/chromium_tests/chromium_fyi.py
[modify] https://crrev.com/73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5/scripts/slave/recipes/chromium.expected/full_chromium_fyi_CrWinAsan.json
[modify] https://crrev.com/73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5/scripts/slave/recipes/chromium.expected/full_chromium_fyi_CrWinAsanCov.json
[modify] https://crrev.com/73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5/scripts/slave/recipes/chromium.expected/full_chromium_fyi_CrWinAsan_dll_.json

The following revision refers to this bug:
  https://chromium.googlesource.com/infra/infra.git/+/34a6ba6f15ad287e5cc819003cf700c75872c645

commit 34a6ba6f15ad287e5cc819003cf700c75872c645
Author: recipe-roller <recipe-roller@chromium.org>
Date: Wed Jun 22 23:35:40 2016

Roll recipe dependencies (trivial).

This is an automated CL created by the recipe roller. This CL rolls recipe
changes from upstream projects (e.g. depot_tools) into downstream projects
(e.g. tools/build).

More info is at https://goo.gl/zkKdpD. Use https://goo.gl/noib3a to file a bug
(or complain)

build:
  https://crrev.com/f75591f68055d86863a3fce9f49534fb8ce86aa9 Deploy build90-b1 (Pixel C) on chromium.gpu.fyi waterfall. (kbr@chromium.org)
  https://crrev.com/73b661438ae50e81c0bdbb7e9d5fe52dae7e4db5 Make CrWinASan* bots true clobber bots. (thakis@chromium.org)

R=kbr@chromium.org,thakis@chromium.org
BUG=622411, 622339 

TBR=martiniss@chromium.org,phajdan.jr@chromium.org

Review-Url: https://codereview.chromium.org/2084303003

[modify] https://crrev.com/34a6ba6f15ad287e5cc819003cf700c75872c645/infra/config/recipes.cfg

Looks like clobbering fixed it:
https://build.chromium.org/p/chromium.fyi/builders/CrWinAsan%28dll%29/builds/4409

I think this was probably caused by David Majnemer's r272960, which made us emit enum types. Enum types are often large, and LLVM's intrinsic enum was the root cause of the same LNK1285 error in LLVM. With r273453 and r273477, we now split up large list records, and things seem to work again.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/chrome/tools/build_limited/scripts/slave/+/56b336e1971b7fb02212492b894104fe71c11fc7

commit 56b336e1971b7fb02212492b894104fe71c11fc7
Author: recipe-roller <recipe-roller@chromium.org>
Date: Thu Jun 23 19:39:29 2016