Linux (and Android): Implement GetDefaultConcurrentLinks() for Linux Builds

Implement GetDefaultConcurrentLinks() for Linux Builds.

This method is used to determine the size of the Ninja pool
which is used for links during the build. It is currently size 1 for Linux, because this method is not implemented.
After this patch it will be <total_memory> / 8Gb.

BUG=
R=scottmg@chromium.org

Committed: https://code.google.com/p/gyp/source/detail?r=1701

[anton1, 2013-08-21 15:28:13]

Add Darin as there doesn't seem to be OWNERS in this repo.

[scottmg, 2013-08-21 15:44:58]

(darin to CC, there's separate committers for gyp)

https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py
File pylib/gyp/generator/ninja.py (right):

https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py#...
pylib/gyp/generator/ninja.py:1487: meminfo = open("/proc/meminfo")
with open("/proc/meminfo") as meminfo:
  ...

https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py#...
pylib/gyp/generator/ninja.py:1494: return max(1, int(match.group(1)) / (8 * (2
** 20)))  # total_Kb / 8GB
change comment to match (MB) or remove

[anton1, 2013-08-21 16:05:41]

On 2013/08/21 15:44:58, scottmg wrote:
> (darin to CC, there's separate committers for gyp)
> 
> https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py
> File pylib/gyp/generator/ninja.py (right):
> 
>
https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py#...
> pylib/gyp/generator/ninja.py:1487: meminfo = open("/proc/meminfo")
> with open("/proc/meminfo") as meminfo:
>   ...
> 
>
https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py#...
> pylib/gyp/generator/ninja.py:1494: return max(1, int(match.group(1)) / (8 * (2
> ** 20)))  # total_Kb / 8GB
> change comment to match (MB) or remove

Please advise as to the correct approach for dealing with the Trybots and CQ.

[anton1, 2013-08-21 16:06:58]

Please advise as to the correct approach for dealing with the Trybots and CQ.

https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py
File pylib/gyp/generator/ninja.py (right):

https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py#...
pylib/gyp/generator/ninja.py:1487: meminfo = open("/proc/meminfo")
On 2013/08/21 15:44:58, scottmg wrote:
> with open("/proc/meminfo") as meminfo:
>   ...

Done.

https://codereview.chromium.org/23242012/diff/1/pylib/gyp/generator/ninja.py#...
pylib/gyp/generator/ninja.py:1494: return max(1, int(match.group(1)) / (8 * (2
** 20)))  # total_Kb / 8GB
On 2013/08/21 15:44:58, scottmg wrote:
> change comment to match (MB) or remove

Done.

[scottmg, 2013-08-21 16:19:53]

lgtm

There's no CQ, I can land it for you.

+iannucci fyi as this will affect bots.

[anton1, 2013-08-21 16:21:29]

On 2013/08/21 16:19:53, scottmg wrote:
> lgtm
> 
> There's no CQ, I can land it for you.

Ok, thanks.

> +iannucci fyi as this will affect bots.

[scottmg, 2013-08-21 17:06:59]

Committed patchset #2 manually as r1701.

[Nico, 2013-08-27 22:01:04]

drive-by nit

https://codereview.chromium.org/23242012/diff/11001/pylib/gyp/generator/ninja.py
File pylib/gyp/generator/ninja.py (right):

https://codereview.chromium.org/23242012/diff/11001/pylib/gyp/generator/ninja...
pylib/gyp/generator/ninja.py:1494: return max(1, int(match.group(1)) / (8 * (2
** 20)))
Isn't 2**20 a megabyte, not a gigabyte?

...ah, /proc/meminfo stores kB, not bytes. This isn't super clear, maybe

  available_memory_bytes = int(match.group(1)) * 1024
  return max(1, available_memory_bytes / (8 * (2 ** 30)))

would be a bit clearer?

[anton1, 2013-08-28 09:31:52]

On 2013/08/27 22:01:04, Nico wrote:
> drive-by nit
> 
>
https://codereview.chromium.org/23242012/diff/11001/pylib/gyp/generator/ninja.py
> File pylib/gyp/generator/ninja.py (right):
> 
>
https://codereview.chromium.org/23242012/diff/11001/pylib/gyp/generator/ninja...
> pylib/gyp/generator/ninja.py:1494: return max(1, int(match.group(1)) / (8 * (2
> ** 20)))
> Isn't 2**20 a megabyte, not a gigabyte?
> 
> ...ah, /proc/meminfo stores kB, not bytes. This isn't super clear, maybe
> 
>   available_memory_bytes = int(match.group(1)) * 1024
>   return max(1, available_memory_bytes / (8 * (2 ** 30)))
> 
> would be a bit clearer?

I removed my comment, which said it was Kb, at Scott's suggestion. As this has
already landed, feel free to make a change yourself.