It seems that after r430603 there is an ambiguity whether content_browsertests refers to //content/test:content_browsertests or to //testing/buildbot/filters:content_browsertests.  This ambiguity doesn't exist for normal linux builds - only for Android.  For example - I can repro with:

$ cat out/andr/args.gn 
target_os = "android"
target_cpu = "x86"
is_debug = false
is_component_build = true
symbol_level = 1
use_goma = true

$ ninja -C out/andr -j 100 -l 15 content_browsertests
ninja: Entering directory `out/andr'
ninja: error: unknown target 'content_browsertests'


But I cannot repro with:

$ cat out/gn/args.gn 
dcheck_always_on = true
is_asan = true
is_component_build = true
is_debug = false
use_goma = true
enable_nacl = false

$ ninja -C out/gn -j 100 -l 15 content_browsertests
ninja: Entering directory `out/gn'
...
[0 + 1 / 1 @ 1.1/s : 0.934s] Regenerating ninja files
[0 + 11 / 11 @ 0.1/s : 75.767s] LINK ./content_browsertests

Oh, and on Android, referring to the full path of the target works fine (i.e. doesn't complain about "unknown target"):

$ cat out/andr/args.gn 
target_os = "android"
target_cpu = "x86"
is_debug = false
is_component_build = true
symbol_level = 1
use_goma = true

$ ninja -C out/andr -j 1500 -l 15 content/test:content_browsertests
ninja: Entering directory `out/andr'
ninja: no work to do.

brettw@, could you please advise on the best next steps here?

One way to disambiguate is to rename //testing/buildbot/filters:content_browsertests
target to //testing/buildbot/filters:content_browsertests_filters (https://codereview.chromium.org/2491553002).  I can easily test this change locally and verify that it does indeed help.  OTOH, I recall that I've seen style advise in some GN docs (cannot find a reference right now :-( ) to avoid adding artificial prefixes or suffixes only to disambiguate targets from different directories.

Another way to disambiguate would to be use a full path in testing/buildbot/chromium.fyi.json in the config for Site Isolation Android bot.  This approach would be more difficult to test locally (i.e. I would have to land and then check if the Site Isolation Android bot got better).  Additionally, Android developers might still run into the ambiguity and be confused why ninja says that content_browsertests is an "unknown target".

WDYT?

I believe that jbudorick@ is voting for option #1 (i.e. https://codereview.chromium.org/2491553002) when commenting at https://codereview.chromium.org/2469353002/#msg47

Yeah, I'd vote for #1. We're seeing this elsewhere, too.

We need a fix (either a revert of the root cause CL or a fix CL) landed by 2 PM PT or else our builds tonight are going to fail too.  Is there a fix patch working its way through the CQ now?  If not, can we please revert the root cause from trunk immediately?

Note that I just reverted on yesterday night's canary branch, but we still need a fix on trunk.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/774c1826afe25f57b10919d8cc8f8f4423bb926a

commit 774c1826afe25f57b10919d8cc8f8f4423bb926a
Author: Alex Mineer <amineer@chromium.org>
Date: Wed Nov 09 20:20:50 2016

Revert "Convert Site Isolation Win bot to --isolate-extensions [2nd attempt]."

This reverts commit e8b43a499acd3df1ff12f08f2652d4acca2695fd.

Reverting on branch 2914 (canary branch) only to ship a build today.

Please ensure the issue is address on trunk ASAP.

BUG=663625, 663785 

Cr-Commit-Position: refs/branch-heads/2914@{#3}
Cr-Branched-From: a081fcfbc7471a6681142d49491b82f7b8402851-refs/heads/master@{#430837}

[modify] https://crrev.com/774c1826afe25f57b10919d8cc8f8f4423bb926a/chrome/test/BUILD.gn
[modify] https://crrev.com/774c1826afe25f57b10919d8cc8f8f4423bb926a/content/test/BUILD.gn
[modify] https://crrev.com/774c1826afe25f57b10919d8cc8f8f4423bb926a/testing/buildbot/chromium.fyi.json
[delete] https://crrev.com/053f6e8e059ceba139a584c213c6cbd046839b8f/testing/buildbot/filters/BUILD.gn

This was fixed by the revert in https://codereview.chromium.org/2488963003.