Add a --check flag. This causes gyp to check syntax of files more carefully,...

Add a --check flag. This causes gyp to check syntax of files more carefully,
but makes it about 50% slower, so defaults to off.

[Ben Laurie (Chromium), 2009-09-11 15:43:34]

On Thu, Sep 10, 2009 at 12:16 AM, Ben Laurie<benl@chromium.org> wrote:
> On Tue, Sep 8, 2009 at 12:21 AM, Mark Mentovai<mark@chromium.org> wrote:
>> Nico suggested:
>>> How about having a dumb preprocessor that replaces all "{" that are
>>> not in strings or comments with "mydictcons([", all ":" with ",", all
>>> "}" with "])" and make mydictcons() a function that creates dicts from
>>> a list, but does duplicate detection? (probably easier than writing a
>>> complete parser for the subset of python that's used in gyp)
>>
>> We can certainly do this in a "gyp lint" checker. =A0To make it part of
>> a mainline GYP run, it'd have to not introduce a speed penalty. =A0I
>> think that even one second added to a GYP run would be too much. =A0We
>> chose the Python eval() for speed.
>
> I wrote an eval/checker in my copious spare time. My Python-fu is
> weak, so it probably sucks (I'd love to know how I _should_ have done
> it), but it does parse gyp files and detect the error I made, and it
> uses Python's own compiler so probably reasonably fast...

http://codereview.chromium.org/202059

>
> Attached...
>
>>
>> Mark
>>
>

[Ben Laurie (Chromium), 2009-09-12 09:49:27]

Did anyone notice this?

[M-A Ruel, 2009-09-18 13:04:55]

Can you get numbers w.r.t. load time incurred by import compiler? I don't know,
maybe it was already imported elsewhere.

http://codereview.chromium.org/202059/diff/4001/4002
File pylib/gyp/__init__.py (right):

http://codereview.chromium.org/202059/diff/4001/4002#newcode33
Line 33: includes=[], depth='.', params={}, check=False):
[Unrelated with this change] Wouha, default mutable parameters. Sooo bad.

http://codereview.chromium.org/202059/diff/4001/4003
File pylib/gyp/input.py (right):

http://codereview.chromium.org/202059/diff/4001/4003#newcode3
Line 3: from compiler.ast import Const
I just wonder if it slows down significantly the script load time. compiler is a
large piece of chunk. :/

[Ben Laurie (Chromium), 2009-09-18 13:17:32]

http://codereview.chromium.org/202059/diff/4001/4002
File pylib/gyp/__init__.py (right):

http://codereview.chromium.org/202059/diff/4001/4002#newcode33
Line 33: includes=[], depth='.', params={}, check=False):
On 2009/09/18 13:04:55, Marc-Antoine Ruel wrote:
> [Unrelated with this change] Wouha, default mutable parameters. Sooo bad.

It is?

http://codereview.chromium.org/202059/diff/4001/4003
File pylib/gyp/input.py (right):

http://codereview.chromium.org/202059/diff/4001/4003#newcode3
Line 3: from compiler.ast import Const
On 2009/09/18 13:04:55, Marc-Antoine Ruel wrote:
> I just wonder if it slows down significantly the script load time. compiler is
a
> large piece of chunk. :/

I commented these out and timed the script. I can't tell the difference with and
without (with --check off, of course).

[M-A Ruel, 2009-09-18 13:21:07]

http://codereview.chromium.org/202059/diff/4001/4003
File pylib/gyp/input.py (right):

http://codereview.chromium.org/202059/diff/4001/4003#newcode119
Line 119: def CheckedEval(file_contents):
Add a description about what it does exactly.

http://codereview.chromium.org/202059/diff/4001/4003#newcode120
Line 120: ast = compiler.parse(file_contents)
alignment is wrong, it should be +2

[Ben Laurie (Chromium), 2009-09-18 14:03:27]

http://codereview.chromium.org/202059/diff/4001/4003
File pylib/gyp/input.py (right):

http://codereview.chromium.org/202059/diff/4001/4003#newcode119
Line 119: def CheckedEval(file_contents):
On 2009/09/18 13:21:07, Marc-Antoine Ruel wrote:
> Add a description about what it does exactly.

Done.

http://codereview.chromium.org/202059/diff/4001/4003#newcode120
Line 120: ast = compiler.parse(file_contents)
On 2009/09/18 13:21:07, Marc-Antoine Ruel wrote:
> alignment is wrong, it should be +2

Done.

[Mark Mentovai, 2009-09-19 18:01:18]

OK, I checked this in along with updated test expectations where it count at
lucky r666.  Thanks for the patch.