Roll Clang 214024:216630 (+216684) and switch to CMake (try 3)

Roll Clang 214024:216630 (+216684) and switch to CMake (try 3)

This updates Chromium's clang version to r216630 with
r216684 cherry-picked to fix an ASan issue.

It also changes the build script for Clang to use CMake
instead of Autoconf. The ASan team say this configuration
is better tested, and it also makes us consistent with
the Windows Clang build which already uses CMake.

BUG=400849
R=thakis@chromium.org
TBR=brettw

Committed: https://crrev.com/bb493574c5f5ca5b572849676b69ab7bfec9828d
Cr-Commit-Position: refs/heads/master@{#295289}

[hans, 2014-09-17 00:26:32]

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

[hans, 2014-09-17 00:26:33]

Try 1: https://codereview.chromium.org/453513004
Try 2: https://codereview.chromium.org/578623002

This one contains a fix to make sure we set RPATH when building the plugins so
they use the bundled stdlibc++/libc++ instead of the system's.

I copied the Mac version of this to the builder that failed the last time, and
it looks good:

With the old 216630:

$ DYLD_PRINT_LIBRARIES=1 bin/clang  -Xclang -load -Xclang
lib/libFindBadConstructs.dylib -c /tmp/a.c
dyld: loaded: /private/tmp/old_clang/bin/clang
dyld: loaded: /usr/lib/libncurses.5.4.dylib
dyld: loaded: /usr/lib/libz.1.dylib
dyld: loaded: /private/tmp/old_clang/bin/libc++.dylib
dyld: loaded: /usr/lib/libSystem.B.dylib
dyld: loaded: /usr/lib/system/libmathCommon.A.dylib
dyld: loaded: /private/tmp/old_clang/bin/clang
dyld: loaded: /usr/lib/libncurses.5.4.dylib
dyld: loaded: /usr/lib/libz.1.dylib
dyld: loaded: /private/tmp/old_clang/bin/libc++.dylib
dyld: loaded: /usr/lib/libSystem.B.dylib
dyld: loaded: /usr/lib/system/libmathCommon.A.dylib
dyld: loaded: /private/tmp/old_clang/lib/libFindBadConstructs.dylib
dyld: unloaded: /private/tmp/old_clang/lib/libFindBadConstructs.dylib
error: unable to load plugin 'lib/libFindBadConstructs.dylib':
'dlopen(lib/libFindBadConstructs.dylib, 9): Library not loaded:
/usr/lib/libc++.1.dylib
  Referenced from: /private/tmp/old_clang/lib/libFindBadConstructs.dylib
  Reason: image not found'


With the new one:

$ DYLD_PRINT_LIBRARIES=1 bin/clang  -Xclang -load -Xclang
lib/libFindBadConstructs.dylib -c /tmp/a.c
dyld: loaded: /private/tmp/new_clang/bin/clang
dyld: loaded: /usr/lib/libncurses.5.4.dylib
dyld: loaded: /usr/lib/libz.1.dylib
dyld: loaded: /private/tmp/new_clang/bin/libc++.dylib
dyld: loaded: /usr/lib/libSystem.B.dylib
dyld: loaded: /usr/lib/system/libmathCommon.A.dylib
dyld: loaded: /private/tmp/new_clang/bin/clang
dyld: loaded: /usr/lib/libncurses.5.4.dylib
dyld: loaded: /usr/lib/libz.1.dylib
dyld: loaded: /private/tmp/new_clang/bin/libc++.dylib
dyld: loaded: /usr/lib/libSystem.B.dylib
dyld: loaded: /usr/lib/system/libmathCommon.A.dylib
dyld: loaded: /private/tmp/new_clang/lib/libFindBadConstructs.dylib

[hans, 2014-09-17 00:40:42]

Whitespace tryjobs for comparison: https://codereview.chromium.org/405473002

[Nico, 2014-09-17 17:13:23]

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

[Nico, 2014-09-17 17:13:23]

lgtm

https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
File tools/clang/scripts/update.sh (right):

https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
tools/clang/scripts/update.sh:306: svn revert
test/asan/TestCases/Linux/new_array_cookie_test.cc
:-)

[hans, 2014-09-17 17:18:52]

Committed patchset #2 (id:20001) manually as bb49357 (presubmit successful).

[commit-bot: I haz the power, 2014-09-17 17:18:52]

Patchset 2 (id:??) landed as
https://crrev.com/bb493574c5f5ca5b572849676b69ab7bfec9828d
Cr-Commit-Position: refs/heads/master@{#295289}

[Alexander Potapenko, 2014-12-17 10:57:51]

On 2014/09/17 17:18:52, I haz the power (commit-bot) wrote:
> Patchset 2 (id:??) landed as
> https://crrev.com/bb493574c5f5ca5b572849676b69ab7bfec9828d
> Cr-Commit-Position: refs/heads/master@{#295289}

By the way, why're you using `strip -x` for ASan dylib on Mac, but only `strip
-g` on Linux?

[hans, 2014-12-17 17:29:30]

On 2014/12/17 10:57:51, Alexander Potapenko wrote:
> On 2014/09/17 17:18:52, I haz the power (commit-bot) wrote:
> > Patchset 2 (id:??) landed as
> > https://crrev.com/bb493574c5f5ca5b572849676b69ab7bfec9828d
> > Cr-Commit-Position: refs/heads/master@{#295289}
> 
> By the way, why're you using `strip -x` for ASan dylib on Mac, but only `strip
> -g` on Linux?

Good question. I think the first discussion about stripping was here:
https://code.google.com/p/chromium/issues/detail?id=256342

Then it seems you added strip -g for Linux in
https://codereview.chromium.org/303473011

[Nico, 2015-04-29 19:58:46]

https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
File tools/clang/scripts/update.sh (right):

https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
tools/clang/scripts/update.sh:242: # We have moved from building compiler-rt in
the LLVM tree, to a separate
(this is annoying every time I need to debug something with compiler-rt :-/)

[hans, 2015-04-29 20:02:33]

https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
File tools/clang/scripts/update.sh (right):

https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
tools/clang/scripts/update.sh:242: # We have moved from building compiler-rt in
the LLVM tree, to a separate
On 2015/04/29 19:58:46, Nico wrote:
> (this is annoying every time I need to debug something with compiler-rt :-/)

The deleting part or the fact that it moved? Are you trying local changes in
third_party/llvm/?

[Nico, 2015-04-29 20:04:54]

On 2015/04/29 20:02:33, hans wrote:
>
https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
> File tools/clang/scripts/update.sh (right):
> 
>
https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
> tools/clang/scripts/update.sh:242: # We have moved from building compiler-rt
in
> the LLVM tree, to a separate
> On 2015/04/29 19:58:46, Nico wrote:
> > (this is annoying every time I need to debug something with compiler-rt :-/)
> 
> The deleting part or the fact that it moved? Are you trying local changes in
> third_party/llvm/?

That it's out of tree. I have my llvm checkout in ~/src/llvm, and if I want to
test asan things I need to go to a separate compiler-rt checkout, build that
manually, carefully copy stuff over, etc. Previously, I could just `svn up` in
~/src/llvm/projects/compiler-rt in the rare case that I needed the runtime.

[Nico, 2015-04-29 20:08:12]

On 2015/04/29 20:04:54, Nico wrote:
> On 2015/04/29 20:02:33, hans wrote:
> >
>
https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
> > File tools/clang/scripts/update.sh (right):
> > 
> >
>
https://codereview.chromium.org/575833003/diff/20001/tools/clang/scripts/upda...
> > tools/clang/scripts/update.sh:242: # We have moved from building compiler-rt
> in
> > the LLVM tree, to a separate
> > On 2015/04/29 19:58:46, Nico wrote:
> > > (this is annoying every time I need to debug something with compiler-rt
:-/)
> > 
> > The deleting part or the fact that it moved? Are you trying local changes in
> > third_party/llvm/?
> 
> That it's out of tree. I have my llvm checkout in ~/src/llvm, and if I want to
> test asan things I need to go to a separate compiler-rt checkout, build that
> manually, carefully copy stuff over, etc. Previously, I could just `svn up` in
> ~/src/llvm/projects/compiler-rt in the rare case that I needed the runtime.

…but now I realize that in my current case, I need an android asan runtime which
was always complicated. So today the new way isn't much worse than the old way,
I suppose.