Trybots on linux_view/linux_chromeos are failing when building NaCl

[Anton Vayvod]

Hi,

trybots for linux_chromeos and linux_view are failing for me with error:

Falling back to local compiler: /usr/bin/g++
g++: native_client/src/shared/platform/linux/condition_variable.cc: No such file or directory
g++: no input files
make: *** [out/Debug/obj.target/platform/native_client/src/shared/platform/linux/condition_variable.o] Error 1
Falling back to local compiler: /usr/bin/g++
g++: native_client/src/shared/platform/linux/lock.cc: No such file or directory
g++: no input files
make: *** [out/Debug/obj.target/platform/native_client/src/shared/platform/linux/lock.o] Error 1

Any thoughts on how to fix this?

Thanks,

Anton.

[Ryan Sleevi]

The discussion earlier today on IRC was to force a clobber build as
part of your try job submission. Something to do with how NACL DEPS
are rolled.

Logs are at http://echelog.matzon.dk/logs/browse/chromium/1305583200 ,
[01:57:53] - [02:04:27] between willchan, evmar, and bbudge.

[Timur Iskhodzhanov]

I'll clobber the FYI waterfall bots then
(failing too)

> --
> Chromium Developers mailing list: chromi...@chromium.org
> View archives, change email options, or unsubscribe:
>    http://groups.google.com/a/chromium.org/group/chromium-dev
>

[Anton Vayvod]

Thanks Ryan!

[02:07:14] <willchan> we should probably email chromium-dev so other linux folks who sync know to clobber?
[02:08:03] <leiz> willchan: sgtm

Occasionally implemented :)

On Tue, May 17, 2011 at 11:53 AM, Ryan Sleevi <rsl...@chromium.org> wrote:
>

[Evan Martin]

On Tue, May 17, 2011 at 12:53 AM, Ryan Sleevi <rsl...@chromium.org> wrote:
>> g++: native_client/src/shared/platform/linux/condition_variable.cc: No such
>> file or directory

Ah, this is an old bug in the gyp makefile generator where it falls
down when you rename the input file in such a way that it produces the
same output. (Here: condition_variable.cc -> condition_variable.c.)

I haven't looked into how hard it would be to solve. You wouldn't
expect it to happen very often...

[Evan Martin]

If you don't want to rebuild from scratch:
rm -rf out/{Debug,Release}/obj.target/platform/native_client
should work around it.

[Randy Smith]

I found that in addition I had to:

rm -f out/Debug/.deps/all.deps
rm -rf out/Debug/.deps/out/Debug/obj.target/platform/native_client/src/shared/platform/linux

No guarantees that that's safe--I don't understand our build system
well enough. But it got me past the platform/native_client problem,
and didn't (I believe) force a full rebuild.

-- Randy

[Evan Martin]

On Tue, May 17, 2011 at 1:39 PM, Randy Smith <rds...@google.com> wrote:
> On Tue, May 17, 2011 at 2:12 PM, Evan Martin <ev...@chromium.org> wrote:
>> On Tue, May 17, 2011 at 11:11 AM, Evan Martin <ev...@chromium.org> wrote:
>>> On Tue, May 17, 2011 at 12:53 AM, Ryan Sleevi <rsl...@chromium.org> wrote:
>>>>> g++: native_client/src/shared/platform/linux/condition_variable.cc: No such
>>>>> file or directory
>>>
>>> Ah, this is an old bug in the gyp makefile generator where it falls
>>> down when you rename the input file in such a way that it produces the
>>> same output.  (Here: condition_variable.cc -> condition_variable.c.)
>>>
>>> I haven't looked into how hard it would be to solve.  You wouldn't
>>> expect it to happen very often...
>>
>> If you don't want to rebuild from scratch:
>>  rm -rf out/{Debug,Release}/obj.target/platform/native_client
>> should work around it.
>
> I found that in addition I had to:
>
> rm -f out/Debug/.deps/all.deps
> rm -rf out/Debug/.deps/out/Debug/obj.target/platform/native_client/src/shared/platform/linux
>
> No guarantees that that's safe--I don't understand our build system
> well enough.  But it got me past the platform/native_client problem,
> and didn't (I believe) force a full rebuild.

Whoops, yeah, I screwed up the above path; I meant:
 rm -rf out/.deps/out/{Debug,Release}/obj.target/platform/native_client