ninja windows: workaround for too long names

ninja windows: workaround for too long names

Hash rule_name/unique_name for now to make the path shorter and get builds working again. We should come up with a better way of doing this in general.

(Jay was running into this problem, and found this to be a reasonable/working workaround) 

BUG=276
TEST=

Committed: https://code.google.com/p/gyp/source/detail?r=1421

[Nico, 2012-06-22 23:47:39]

https://chromiumcodereview.appspot.com/10665008/diff/4001/pylib/gyp/generator...
File pylib/gyp/generator/ninja.py (right):

https://chromiumcodereview.appspot.com/10665008/diff/4001/pylib/gyp/generator...
pylib/gyp/generator/ninja.py:547: name = '%s_%s' % (self.qualified_target,
rule['rule_name'])
Could we use target_ + hash(qualified_target) here instead? Then it looks
slightly less ugh-inducing.

Longer-term, steveblock can come up with a patch that omits the hash suffix if
just the target name is unique already.

[scottmg, 2012-06-22 23:51:09]

https://chromiumcodereview.appspot.com/10665008/diff/4001/pylib/gyp/generator...
File pylib/gyp/generator/ninja.py (right):

https://chromiumcodereview.appspot.com/10665008/diff/4001/pylib/gyp/generator...
pylib/gyp/generator/ninja.py:547: name = '%s_%s' % (self.qualified_target,
rule['rule_name'])
On 2012/06/22 23:47:39, Nico wrote:
> Could we use target_ + hash(qualified_target) here instead? Then it looks
> slightly less ugh-inducing.
> 
> Longer-term, steveblock can come up with a patch that omits the hash suffix if
> just the target name is unique already.

sgtm, let me give it a test.

[Nico, 2012-06-22 23:52:16]

https://chromiumcodereview.appspot.com/10665008/diff/4001/pylib/gyp/generator...
File pylib/gyp/generator/ninja.py (right):

https://chromiumcodereview.appspot.com/10665008/diff/4001/pylib/gyp/generator...
pylib/gyp/generator/ninja.py:547: name = '%s_%s' % (self.qualified_target,
rule['rule_name'])
On 2012/06/22 23:51:09, scottmg wrote:
> On 2012/06/22 23:47:39, Nico wrote:
> > Could we use target_ + hash(qualified_target) here instead? Then it looks
> > slightly less ugh-inducing.
> > 
> > Longer-term, steveblock can come up with a patch that omits the hash suffix
if
> > just the target name is unique already.
> 
> sgtm, let me give it a test.

Cool. For actions too, and if qualified_target is used elsewhere, there too :-)

[scottmg, 2012-06-22 23:57:16]

oops! :( i think i caught 'em all now.

[Nico, 2012-06-23 00:00:06]

lgtm!

Thanks for fixing this so quickly.

https://chromiumcodereview.appspot.com/10665008/diff/7002/pylib/gyp/generator...
File pylib/gyp/generator/ninja.py (right):

https://chromiumcodereview.appspot.com/10665008/diff/7002/pylib/gyp/generator...
pylib/gyp/generator/ninja.py:514: name = '%s_%s' %
(hashlib.md5(self.qualified_target).hexdigest(),
nit: I think it'd look slightly nicer if you swapped the tuple elements
("do_stuff_ba5f0000000" instead of "ba8f00000000_do_stuff"). Also below. But not
important, it's just a workaround after all.

[jaysoffian, 2012-06-23 00:04:21]

On 2012/06/22 23:57:16, scottmg wrote:
> oops! :( i think i caught 'em all now.

This patch works for me. The longest entry is now:

concatenated_script_formatter_worker_js_target_41e3628345045b3afe455371f5a4b92c_concatenate_script_formatter_worker_js.$unique_name.rsp

(and $unique_name isn't defined for that particular rule.)

[scottmg, 2012-06-23 00:05:37]

On 2012/06/23 00:04:21, jaysoffian wrote:
> On 2012/06/22 23:57:16, scottmg wrote:
> > oops! :( i think i caught 'em all now.
> 
> This patch works for me. The longest entry is now:

Thanks for your help!

[scottmg, 2012-06-23 00:05:52]

On 2012/06/23 00:00:06, Nico wrote:
> lgtm!
> 
> Thanks for fixing this so quickly.
> 
>
https://chromiumcodereview.appspot.com/10665008/diff/7002/pylib/gyp/generator...
> File pylib/gyp/generator/ninja.py (right):
> 
>
https://chromiumcodereview.appspot.com/10665008/diff/7002/pylib/gyp/generator...
> pylib/gyp/generator/ninja.py:514: name = '%s_%s' %
> (hashlib.md5(self.qualified_target).hexdigest(),
> nit: I think it'd look slightly nicer if you swapped the tuple elements
> ("do_stuff_ba5f0000000" instead of "ba8f00000000_do_stuff"). Also below. But
not
> important, it's just a workaround after all.

Done.

[Steve Block, 2012-06-25 11:40:06]

Thanks Scott!