mac: Use ninja by default.

mac: Use ninja by default.

With this change, `gclient runhooks` creates ninja files by default on Mac.
To build, run `ninja -C out/Release chrome`. The output appears in out/
instead of in xcodebuild/. (This doesn't happen if OS=ios, so it shouldn't
affect chrome/ios development.)

If you need to use xcode for some reason, you can
`export GYP_GENERATORS=xcode` to force xcode project generation.
If you do this, please inform me why.

Bots should automatically pick up the binaries in the new location (out).
If something does break, it should be safe to revert this change, and bots
should then automatically go back to picking up binaries from the old
location (xcodebuild).

BUG=294387
R=mark@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=231675

[Nico, 2013-10-24 23:53:16]

How can I test the installer bits?

[Mark Mentovai, 2013-10-25 12:58:18]

LGTM but I suggest making the recommended additions.

Testing info: you can test the installer bits by building the stuff in
chrome/installer/mac/third_party (goobsdiff, goobspatch, xz, and xzdec). The
dirdiffer and dirpatcher scripts use those tools. dirdiffer actually runs
dirpatcher for a verification step, so all you need to do is run dirdiffer on a
pair of directories and see that it succeeds (exit status 0). You can run
dirdiffer by running "dirdiffer.sh old_dir new_dir patch_dir". In real life, it
would run on, say, the Chromium versioned directories, so you could do
"dirdiffer.sh path/to/OldChrome.app/Contents/Versions/32.0.1665.0
path/to/NewChrome.app/Contents/Versions/32.0.1666.0 /tmp/1665-to-1666-patch".
Because Chrome is huge, dirdiffer takes a long time, like a half hour.
dirpatcher is much quicker, like under a minute.

https://codereview.chromium.org/31143002/diff/30001/chrome/installer/mac/dird...
File chrome/installer/mac/dirdiffer.sh (right):

https://codereview.chromium.org/31143002/diff/30001/chrome/installer/mac/dird...
chrome/installer/mac/dirdiffer.sh:129:
tool="${BASH_REMATCH[1]}/src/out/Release/${tool_name}"
You can make this check the ninja locations, and if it’s not found there, then
the xcode locations.

https://codereview.chromium.org/31143002/diff/30001/chrome/installer/mac/dirp...
File chrome/installer/mac/dirpatcher.sh (right):

https://codereview.chromium.org/31143002/diff/30001/chrome/installer/mac/dirp...
chrome/installer/mac/dirpatcher.sh:72:
tool="${BASH_REMATCH[1]}/src/out/Release/${tool_name}"
Same here.

[Nico, 2013-10-25 14:14:44]

Thanks!

https://codereview.chromium.org/31143002/diff/30001/chrome/installer/mac/dird...
File chrome/installer/mac/dirdiffer.sh (right):

https://codereview.chromium.org/31143002/diff/30001/chrome/installer/mac/dird...
chrome/installer/mac/dirdiffer.sh:129:
tool="${BASH_REMATCH[1]}/src/out/Release/${tool_name}"
On 2013/10/25 12:58:19, Mark Mentovai wrote:
> You can make this check the ninja locations, and if it’s not found there, then
> the xcode locations.

I had this at first (see diff from patch set 1 to 2 to see how I incompletely
reverted that change), but then I realized that this is changing in the same CL
as the generator default – so if this CL is reverted, both the generator change
and this file get reverted in the same revision. So I think just using the
canonical location here should be enough.

[Mark Mentovai, 2013-10-25 14:29:49]

You’re just changing the generator default, not making Xcode not work at all. I
don’t want this to break if someone decides to go back to Xcode on a bot by
changing the generator it uses in the bot config or something instead of backing
this change out.

[Nico, 2013-10-25 14:46:30]

I'm not making xcode not work at all, but the point is to not use it on
bots that build installers (or any bots, really). If that becomes necessary
for some reason, we can make this script more complicated then, right?
On Oct 25, 2013 7:29 AM, <mark@chromium.org> wrote:

> You’re just changing the generator default, not making Xcode not work at
> all. I
> don’t want this to break if someone decides to go back to Xcode on a bot by
> changing the generator it uses in the bot config or something instead of
> backing
> this change out.
>
>
https://codereview.chromium.**org/31143002/<https://codereview.chromium.org/3...
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[Mark Mentovai, 2013-10-25 15:23:19]

I don’t agree. Since the generator switch exists, someone’s bound to touch it at
some point, maybe not for Chrome, but maybe for some other project in the Chrome
universe that shares these scripts and needs to keep using Xcode for some
reason. Things that we can identify as needing to be robust against such switch
flips ahead of time should be beefed up. No sense in causing
difficult-to-troubleshoot problems for someone else down the line.

[Nico, 2013-10-25 15:29:14]

Are you thinking of any project in particular?

This whole switch was a lot more complicated than it could've been due to
many places supported flexibility that was never used in practice. Since I
just lost several days because of that, I don't want to add it here.
On Oct 25, 2013 8:23 AM, <mark@chromium.org> wrote:

> I don’t agree. Since the generator switch exists, someone’s bound to touch
> it at
> some point, maybe not for Chrome, but maybe for some other project in the
> Chrome
> universe that shares these scripts and needs to keep using Xcode for some
> reason. Things that we can identify as needing to be robust against such
> switch
> flips ahead of time should be beefed up. No sense in causing
> difficult-to-troubleshoot problems for someone else down the line.
>
>
https://codereview.chromium.**org/31143002/<https://codereview.chromium.org/3...
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[Mark Mentovai, 2013-10-25 15:32:34]

You obviously feel very strongly about this and I have 450 other threads to get
through so LGTM.

[Nico, 2013-10-30 00:04:28]

Committed patchset #2 manually as r231675 (presubmit successful).