(No comment was entered for this change.)

(No comment was entered for this change.)

Removing the mstone:Linux label which is no longer used.  The new 
mstone:LinuxBeta label indicates issues which should be resolved before the 
release of the Linux build to the beta channel.  There is also a 
mstone:LinuxDev label for issues that should be resolved before the release 
of the build to testers and then the dev channel.

The initial assignment was based on the priority.  The entire list should 
be triaged to insure things are in the right bucket.

(No comment was entered for this change.)

Is the startup_tests an appropriate benchmark for this? The article mentioned
optimizing linking can be expensive, so perhaps it would be something we only enable
on official builds.

Yeah, I think if anything this should affect startup.

(We need a test that also involves dropping the disk cache before running, since 
otherwise we're just testing in-memory perf...)

http://ridiculousfish.com/blog/archives/2009/06/22/i-didnt-order-that-so-why-is-it-on-
my-bill-episode-1/ discusses -fvisibility-inlines-hidden

Can you look into this, please? Thanks

Unassigning because assigning to thestig didn't really make sense.

In my unscientific StartupTest.PerfCold test, -Wl,-O1 and -Wl,--hash-style=gnu were
comparable, both about 10% faster.

Did they bloat the executable any?
If not, let's pick one, check it in, and see how the official perf bot likes it...

Rearranging bugs to eliminate LinuxBeta label.

Demoting bugs that wouldn't block a beta release.

Moving this out of mstone-4 and into mstone-5 after talking to the Linux 
team.

(No comment was entered for this change.)

(No comment was entered for this change.)

(No comment was entered for this change.)

Switching from -O2 to -O3 in common.gypi grew the stripped size of the 'chrome' binary by 4%, which doesn't 
seem too bad.  I haven't checked the perf impact yet.

The original article discusses a few different things.

-Wl,O1: this seems to help in my local testing, http://i.imgur.com/VvQsP.png
I committed the change.

--hash-style=gnu: this seems to be the default on the gccs I could find (hardy and later -- well, hardy does hash-style=both which includes gnu-style AFAICT).

I later posted about -fvisibility-inlines-hidden, which is already on in the tree.

That covers everything reported here.  I think in the future I should make bugs more specific, like "see if hash-style=gnu helps", to make them easier to close.

(No comment was entered for this change.)

This issue has been closed for some time. No one will pay attention to new comments.
If you are seeing this bug or have new data, please click New Issue to start a new bug.

(No comment was entered for this change.)