fix: https://codereview.chromium.org/2119653002/

Update: ignore that CL.

I think all you need to do is set 'custom_toolchain = "//build/toolchain/cros:target"' in your GN_ARGS, which you should be doing unconditionally (i.e., regardless of whether you're using clang or not). I always meant for you to be doing this, but forgot to tell you in the midst of getting all of the other stuff working.

Assuming we can confirm that that works, let me know when you have set that, and I can clean up some other things.

 Issue 626058  has been merged into this issue.

This is affecting the chromeos ASAN builders, so we need this sooner than later:

https://build.chromium.org/p/chromiumos.chromium/builders/x86-generic-tot-asan-informational/builds/11156

setting custom_toolchain as in #2 seems to fix the simple chrome workflow. I am testing that. 
However when trying this in the ebuild, I get the error about RPATH which was fixed a couple of weeks ago. 
The patches needed to reach this point are:

https://chromium-review.googlesource.com/#/c/358921/
https://chromium-review.googlesource.com/#/c/358770/

and the failure can be seen in this try server 

ttps://uberchromegw.corp.google.com/i/chromiumos.tryserver/builders/asan/builds/86

follow the build packages stdio link and ignore the "chaps" failures.

The problem as reported:

chromeos-chrome-54.0.2790.0_rc-r1: scanelf: rpath_security_checks(): Security problem with DT_RPATH='$ORIGIN/.' in /build/amd64-generic/tmp/portage/chromeos-base/chromeos-chrome-54.0.2790.0_rc-r1/image/opt/google/chrome/chrome-sandbox with mode set of 4755
chromeos-chrome-54.0.2790.0_rc-r1: scanelf: rpath_security_checks(): Security problem with DT_RPATH='$ORIGIN/.' in /build/amd64-generic/tmp/portage/chromeos-base/chromeos-chrome-54.0.2790.0_rc-r1/image/opt/google/chrome/chrome-sandbox with mode set of 4755
chromeos-chrome-54.0.2790.0_rc-r1: 
chromeos-chrome-54.0.2790.0_rc-r1:  * QA Notice: The following files contain insecure RUNPATHs
chromeos-chrome-54.0.2790.0_rc-r1:  *  Please file a bug about this at http://bugs.gentoo.org/
chromeos-chrome-54.0.2790.0_rc-r1:  *  with the maintaining herd of the package.
chromeos-chrome-54.0.2790.0_rc-r1:  * $ORIGIN/. /build/amd64-generic/tmp/portage/chromeos-base/chromeos-chrome-54.0.2790.0_rc-r1/image/opt/google/chrome/chrome-sandbox
chromeos-chrome-54.0.2790.0_rc-r1: 
chromeos-chrome-54.0.2790.0_rc-r1:  * ERROR: chromeos-base/chromeos-chrome-54.0.2790.0_rc-r1::chromiumos failed:
chromeos-chrome-54.0.2790.0_rc-r1:  *   Aborting due to serious QA concerns with RUNPATH/RPATH
chromeos-chrome-54.0.2790.0_rc-r1:  * 
chromeos-chrome-54.0.2790.0_rc-r1:  * Call stack:
chromeos-chrome-54.0.2790.0_rc-r1:  *   misc-functions.sh, line 1347:  Called install_qa_check
chromeos-chrome-54.0.2790.0_rc-r1:  *   misc-functions.sh, line  445:  Called die
chromeos-chrome-54.0.2790.0_rc-r1:  * The specific snippet of code:
chromeos-chrome-54.0.2790.0_rc-r1:  *   			die "Aborting due to serious QA concerns with RUNPATH/RPATH"
chromeos-chrome-54.0.2790.0_rc-r1:  * 

to summarize, we currently have the following problems:

- chromium:626240
- chromium:626241
- RPATH issue described above

And 2 fixes in flight:

https://chromium-review.googlesource.com/#/c/358921/
https://chromium-review.googlesource.com/#/c/358770/

I think at this point is better if dpranke owns this issue since the 2 unresolved problems are on his side.

Okay, in this case the problem w/ rpath is that, as the GN build is written, chrome_sandbox is linked w/ the sanitizer libraries, and so they need rpath to be set.

The GYP build explicitly turns off the sanitizer libraries (and flags) for the binary.

Arguably, we could just turn off the RPATH check, since it's not like we're going to actually ship the sandbox binary built w/ ASAN, but I'd rather not have to even make that argument, so I'll see if I can turn off the sanitizer configs instead.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/overlays/chromiumos-overlay/+/83a4db66297839d765616356e7bc34e18473a73a

commit 83a4db66297839d765616356e7bc34e18473a73a
Author: Luis Lozano <llozano@chromium.org>
Date: Thu Jul 07 21:51:12 2016

GN: Fix GN builds so that we can use our own version of Clang/LLVM.

Need to set "custom_toolchain" GN var for this to work.

BUG= chromium:624901 
TEST=cbuildbot falco-release daisy-release amd64-generic-asan

Change-Id: Id734f20f55619dbdebba7ee97918aa71b737fb4a
Reviewed-on: https://chromium-review.googlesource.com/358921
Commit-Ready: Luis Lozano <llozano@chromium.org>
Tested-by: Luis Lozano <llozano@chromium.org>
Reviewed-by: Steven Bennetts <stevenjb@chromium.org>
Reviewed-by: Dirk Pranke <dpranke@chromium.org>

[modify] https://crrev.com/83a4db66297839d765616356e7bc34e18473a73a/chromeos-base/chromeos-chrome/chromeos-chrome-9999.ebuild

Should this be fixed now, or is there anything else that needs to be done?

The problem listed in comment #5 is still an issue. I should have a fix for it later today.

OK, cheers!

please dont forget 626240 and 626241 (marked as dependencies for this bug)

Proposed fix: https://codereview.chromium.org/2152153002/

(I posted fixes for the other two issues as well, so hopefully we can unblock things soon).

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/651705fee413b0e84bddbf65ae563c7c6e7e2868

commit 651705fee413b0e84bddbf65ae563c7c6e7e2868
Author: dpranke <dpranke@chromium.org>
Date: Fri Jul 15 03:58:47 2016

Fix rpath dependencies in ASAN builds of chrome_sandbox.

chrome_sandbox is a setuid binary, so we don't want it to pick
up -Wl,-rpath=$ORIGIN. Binaries built with sanitizers might
need that flag in order to pick up the instrumented libraries
properly, so the fix for this is to just not build chrome_sandbox
with any of the sanitizer flags.

R=rickyz@chromium.org, llozano@chromium.org
BUG= 624901 

Review-Url: https://codereview.chromium.org/2152153002
Cr-Commit-Position: refs/heads/master@{#405692}

[modify] https://crrev.com/651705fee413b0e84bddbf65ae563c7c6e7e2868/sandbox/linux/BUILD.gn

Looks like the chromeos ASAN builders cycled green, hooray! Thanks!

I think all of the known issues on this bug have landed (there is still also  bug 618346 ).

I'm going to bounce this back to llozano at this point. I really shouldn't be the owner of "make clang work", so maybe we can use sub-bugs instead next time?

I think the issues reported in this bug have been fixed. 

 Issue 628384  has been merged into this issue.