Use -fdiagnostics-absolute-paths in Win-Clang builds

Use -fdiagnostics-absolute-paths in Win-Clang builds

BUG=636109
Committed: https://crrev.com/be2fa1a7f40f37fb052265b2183b65b3818c6682
Cr-Commit-Position: refs/heads/master@{#414856}

[hans, 2016-08-26 16:08:03]

hans@chromium.org changed reviewers:
+ thakis@chromium.org

[hans, 2016-08-26 16:08:03]

-fdiagnostics-absolute-paths is in Clang r279827

Please take a look.

[Nico, 2016-08-26 16:13:02]

(not sure)

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD.gn
File build/config/compiler/BUILD.gn (right):

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD...
build/config/compiler/BUILD.gn:377: if (is_clang && is_win &&
llvm_force_head_revision) {
Should we do this on all platforms? Seems a bit arbitrary to do this on Windows
but not elsewhere.

[hans, 2016-08-26 16:21:43]

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD.gn
File build/config/compiler/BUILD.gn (right):

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD...
build/config/compiler/BUILD.gn:377: if (is_clang && is_win &&
llvm_force_head_revision) {
On 2016/08/26 16:13:01, Nico wrote:
> Should we do this on all platforms? Seems a bit arbitrary to do this on
Windows
> but not elsewhere.

I agree it might seem a bit odd to do it on only one platform, but I think the
desire for this is very Windows-specific.

Personally I prefer relative paths, and I think this is very natural to most
Linux and Mac folks too.

I guess what it boils down to is that there's no precedence for absolute paths
in diagnostics on Unix (GCC doesn't do it), but there is on Windows and those
developers seem to expect it.

[Nico, 2016-08-26 16:29:22]

+scott, bruce who requested this. Can you chime in on the discussion upthread?

[scottmg, 2016-08-26 16:58:33]

scottmg@chromium.org changed reviewers:
+ scottmg@chromium.org

[scottmg, 2016-08-26 16:58:34]

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD.gn
File build/config/compiler/BUILD.gn (right):

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD...
build/config/compiler/BUILD.gn:377: if (is_clang && is_win &&
llvm_force_head_revision) {
On 2016/08/26 16:21:43, hans wrote:
> On 2016/08/26 16:13:01, Nico wrote:
> > Should we do this on all platforms? Seems a bit arbitrary to do this on
> Windows
> > but not elsewhere.
> 
> I agree it might seem a bit odd to do it on only one platform, but I think the
> desire for this is very Windows-specific.
> 
> Personally I prefer relative paths, and I think this is very natural to most
> Linux and Mac folks too.
> 
> I guess what it boils down to is that there's no precedence for absolute paths
> in diagnostics on Unix (GCC doesn't do it), but there is on Windows and those
> developers seem to expect it.

Yeah, I would agree with keeping it Windows-only. Even though it's a bit odd, 
with /FC in gyp/gn it's been different on both platforms for a long time.

(We have /FC on when running like this, right? Does this also need to be turned
on explicitly still for clang-cl?)

[hans, 2016-08-26 17:02:39]

On 2016/08/26 16:58:34, scottmg wrote:
> (We have /FC on when running like this, right? Does this also need to be
turned
> on explicitly still for clang-cl?)

We decided not to implement /FC in clang-cl as that also enables absolute paths
for __FILE__ directives, which makes builds less hermetic. And instead of just
implementing the diagnostics part of /FC, which would seem confusing, we just
made a separate -fdiagnostics-absolute-paths flag.

[Nico, 2016-08-26 17:06:25]

ok, lgtm

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD.gn
File build/config/compiler/BUILD.gn (right):

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD...
build/config/compiler/BUILD.gn:375: # Windows devs want absolute paths in
diagnostics (crbug.com/636109).
Maybe you can expand this a bit based on discussion ("why only windows?)"

[hans, 2016-08-26 20:50:06]

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD.gn
File build/config/compiler/BUILD.gn (right):

https://codereview.chromium.org/2281963002/diff/1/build/config/compiler/BUILD...
build/config/compiler/BUILD.gn:375: # Windows devs want absolute paths in
diagnostics (crbug.com/636109).
On 2016/08/26 17:06:24, Nico wrote:
> Maybe you can expand this a bit based on discussion ("why only windows?)"

Done.

[hans, 2016-08-26 20:50:21]

The CQ bit was checked by hans@chromium.org

[hans, 2016-08-26 20:50:22]

The patchset sent to the CQ was uploaded after l-g-t-m from thakis@chromium.org
Link to the patchset: https://codereview.chromium.org/2281963002/#ps20001
(title: "expand comment")

[commit-bot: I haz the power, 2016-08-26 20:51:25]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-08-26 23:26:46]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-08-26 23:28:17]

Description was changed from

==========
Use -fdiagnostics-absolute-paths in Win-Clang builds

BUG=636109
==========

to

==========
Use -fdiagnostics-absolute-paths in Win-Clang builds

BUG=636109

Committed: https://crrev.com/be2fa1a7f40f37fb052265b2183b65b3818c6682
Cr-Commit-Position: refs/heads/master@{#414856}
==========

[commit-bot: I haz the power, 2016-08-26 23:28:18]

Patchset 2 (id:??) landed as
https://crrev.com/be2fa1a7f40f37fb052265b2183b65b3818c6682
Cr-Commit-Position: refs/heads/master@{#414856}