Support -malign-double option

Support -malign-double option

While LLVM supports the `-malign-double` option, Clang does not.
Passing in `-mllvm -malign-double` cannot be used as a workaround
as it will result in different machine description being used
by Clang and LLVM.

This change adds support for `-malign-double` option directly into
Clang and sets the machine description accordingly.

BUG=none
R=dschuff@chromium.org

Committed: 88005a1e1adeff1df374266820b61c6304e2e124

[Petr Hosek, 2015-09-02 16:46:23]

phosek@chromium.org changed reviewers:
+ dschuff@chromium.org

[Derek Schuff, 2015-09-02 20:16:23]

LGTM. can you fix the commit message (it's malign-double) and add some detail to
it, and also add LOCALMOD markers?

[Petr Hosek, 2015-09-02 22:11:08]

Patchset #2 (id:20001) has been deleted

[Petr Hosek, 2015-09-02 22:16:13]

Done.

[Derek Schuff, 2015-09-02 22:22:16]

On 2015/09/02 22:16:13, Petr Hosek wrote:
> Done.

oh wait, add a test, in test/Driver

[Petr Hosek, 2015-09-02 23:49:33]

Test added.

[Derek Schuff, 2015-09-03 17:22:05]

https://codereview.chromium.org/1310173005/diff/60001/test/Driver/malign-doub...
File test/Driver/malign-double.c (right):

https://codereview.chromium.org/1310173005/diff/60001/test/Driver/malign-doub...
test/Driver/malign-double.c:1: // RUN: %clang -### -c -save-temps -malign-double
%s \
do we actually need -save-temps in the RUN line?

https://codereview.chromium.org/1310173005/diff/60001/test/Driver/malign-doub...
test/Driver/malign-double.c:2: // RUN:     -target i686-unknown-nacl 2>&1 \
do we actually need to capture stderr too?

[Petr Hosek, 2015-09-03 20:44:02]

I have also added a second (negative) test case.

https://codereview.chromium.org/1310173005/diff/60001/test/Driver/malign-doub...
File test/Driver/malign-double.c (right):

https://codereview.chromium.org/1310173005/diff/60001/test/Driver/malign-doub...
test/Driver/malign-double.c:1: // RUN: %clang -### -c -save-temps -malign-double
%s \
On 2015/09/03 17:22:05, Derek Schuff wrote:
> do we actually need -save-temps in the RUN line?

We don't, removed.

https://codereview.chromium.org/1310173005/diff/60001/test/Driver/malign-doub...
test/Driver/malign-double.c:2: // RUN:     -target i686-unknown-nacl 2>&1 \
On 2015/09/03 17:22:05, Derek Schuff wrote:
> do we actually need to capture stderr too?

We do because clang prints the command being run into stderr rather than stdout.

[Derek Schuff, 2015-09-03 21:16:50]

negative test looks good too, LGTM

[Petr Hosek, 2015-09-04 18:44:08]

Committed patchset #4 (id:80001) manually as
88005a1e1adeff1df374266820b61c6304e2e124 (presubmit successful).