mac: Add the flag "-gline-tables-only" to reduce dSYM size.

mac: Add the flag "-gline-tables-only" to reduce dSYM size.

The dSYM for Google Chrome Framework has grown to be larger than 4GB. dsymutil
can't handle this. Reduce the amount of debug info output for official builds
of Google Chrome. On my local machine, the dSYM is 10x smaller.

BUG=479841
Committed: https://crrev.com/c4fc98d94448349b8fcf3d0627a166c74caf7053
Cr-Commit-Position: refs/heads/master@{#326445}

[erikchen, 2015-04-23 01:01:46]

erikchen@chromium.org changed reviewers:
+ mark@chromium.org, rsesek@chromium.org

[erikchen, 2015-04-23 01:01:47]

rsesek: Please review.

mark@ mentioned he would be OOO for a couple of days. This was the solution he
proposed at https://code.google.com/p/chromium/issues/detail?id=479841#c15

[Nico, 2015-04-23 01:08:53]

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

[Nico, 2015-04-23 01:08:53]

This makes it impossible to debug. Please only do this for official builds.

[Mark Mentovai, 2015-04-23 01:13:49]

We should also work out whether there are any conditions where dsymutil does the
wrong thing. That’s why I asked what versions of the tools were on that bot.
It’s possible that this was a bug in dsymutil that’s been fixed. If that’s truly
the case, then maybe we can get by just by updating the bots.

[Mark Mentovai, 2015-04-23 01:14:50]

Sorry, you answered part of that on the bug, but I saw it LIFO.

[erikchen, 2015-04-23 01:20:36]

thakis: Done.

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi#newco...
build/common.gypi:5135: 'OTHER_CFLAGS': [
I don't understand GYP. I figured that this would override the previous
declaration of OTHER_CFLAGS, but it appears to do the right thing.

[Nico, 2015-04-23 01:26:24]

Looks better, but I'm afraid I have one more bike to shed:

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi#newco...
build/common.gypi:5135: 'OTHER_CFLAGS': [
On 2015/04/23 01:20:36, erikchen wrote:
> I don't understand GYP. I figured that this would override the previous
> declaration of OTHER_CFLAGS, but it appears to do the right thing.

gyp merges lists by appending them. This is documented at
https://code.google.com/p/gyp/wiki/InputFormatReference#Merge_Basics_(=,_?,_+)

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi#newco...
build/common.gypi:5208: 'DEBUG_INFORMATION_FORMAT': 'dwarf-with-dsym',
Probably best to put this here, next to the dsymutil bits?

[Mark Mentovai, 2015-04-23 01:38:06]

https://codereview.chromium.org/1085843004/diff/40001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1085843004/diff/40001/build/common.gypi#newco...
build/common.gypi:5136: # The Google Chrome Framework dSYM generated by dsymutil
has grown
80 columns.

[erikchen, 2015-04-23 01:56:05]

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi#newco...
build/common.gypi:5208: 'DEBUG_INFORMATION_FORMAT': 'dwarf-with-dsym',
On 2015/04/23 01:26:24, Nico (OOO sick) wrote:
> Probably best to put this here, next to the dsymutil bits?

This is a good suggestion, not bikeshedding. Thanks! This reduces the number of
instances of "-gline-tables-only" in ninja files from 2k+ to ~150.

https://codereview.chromium.org/1085843004/diff/40001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1085843004/diff/40001/build/common.gypi#newco...
build/common.gypi:5136: # The Google Chrome Framework dSYM generated by dsymutil
has grown
On 2015/04/23 01:38:06, Mark Mentovai wrote:
> 80 columns.

Done.

[erikchen, 2015-04-23 01:57:49]

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi#newco...
build/common.gypi:5135: 'OTHER_CFLAGS': [
On 2015/04/23 01:26:24, Nico (OOO sick) wrote:
> On 2015/04/23 01:20:36, erikchen wrote:
> > I don't understand GYP. I figured that this would override the previous
> > declaration of OTHER_CFLAGS, but it appears to do the right thing.
> 
> gyp merges lists by appending them. This is documented at
> https://code.google.com/p/gyp/wiki/InputFormatReference#Merge_Basics_(=,_?,_+)

I see. It was unclear to me that lists got merged by default.

[Nico, 2015-04-23 02:16:27]

On 2015/04/23 01:57:49, erikchen wrote:
> https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi
> File build/common.gypi (right):
> 
>
https://codereview.chromium.org/1085843004/diff/20001/build/common.gypi#newco...
> build/common.gypi:5135: 'OTHER_CFLAGS': [
> On 2015/04/23 01:26:24, Nico (OOO sick) wrote:
> > On 2015/04/23 01:20:36, erikchen wrote:
> > > I don't understand GYP. I figured that this would override the previous
> > > declaration of OTHER_CFLAGS, but it appears to do the right thing.
> > 
> > gyp merges lists by appending them. This is documented at
> >
https://code.google.com/p/gyp/wiki/InputFormatReference#Merge_Basics_(=,_?,_+)
> 
> I see. It was unclear to me that lists got merged by default.

Lgtm

[erikchen, 2015-04-23 02:18:50]

The CQ bit was checked by erikchen@chromium.org

[commit-bot: I haz the power, 2015-04-23 02:19:10]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1085843004/60001

[Mark Mentovai, 2015-04-23 02:20:09]

LGTM.

Feel better, Nico!

[commit-bot: I haz the power, 2015-04-23 04:07:48]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-04-23 04:08:49]

Patchset 4 (id:??) landed as
https://crrev.com/c4fc98d94448349b8fcf3d0627a166c74caf7053
Cr-Commit-Position: refs/heads/master@{#326445}

[erikchen, 2015-04-23 16:58:09]

On 2015/04/23 04:08:49, I haz the power (commit-bot) wrote:
> Patchset 4 (id:??) landed as
> https://crrev.com/c4fc98d94448349b8fcf3d0627a166c74caf7053
> Cr-Commit-Position: refs/heads/master@{#326445}

I started a build last night with patch set 4 - it doesn't actually reduce the
binary size. Patch set 3 does reduce the binary size.

[erikchen, 2015-04-23 17:00:24]

On 2015/04/23 16:58:09, erikchen wrote:
> On 2015/04/23 04:08:49, I haz the power (commit-bot) wrote:
> > Patchset 4 (id:??) landed as
> > https://crrev.com/c4fc98d94448349b8fcf3d0627a166c74caf7053
> > Cr-Commit-Position: refs/heads/master@{#326445}
> 
> I started a build last night with patch set 4 - it doesn't actually reduce the
> binary size. Patch set 3 does reduce the binary size.

My guess is that mac_strip!=0 in the conditional is causing the
"gline-tables-only" to not show up. Looking further.

[erikchen, 2015-04-23 18:28:12]

A revert of this CL (patchset #4 id:60001) has been created in
https://codereview.chromium.org/1105703002/ by erikchen@chromium.org.

The reason for reverting is: Reverting this CL because it doesn't actually
reduce the dSYM size generated by dsymutil. .

[erikchen, 2015-04-23 19:23:30]

On 2015/04/23 18:28:12, erikchen wrote:
> A revert of this CL (patchset #4 id:60001) has been created in
> https://codereview.chromium.org/1105703002/ by mailto:erikchen@chromium.org.
> 
> The reason for reverting is: Reverting this CL because it doesn't actually
> reduce the dSYM size generated by dsymutil. .

Through experimentation, it appears that two separate conditionals prevent the
binary size from getting reduced.

'mac_real_dsym == 1'

Despite its name, mac_real_dsym is only true for ASAN builds. Yay.

['(_type=="executable" or _type=="shared_library" or \
    _type=="loadable_module")', {