Introduce '--no-duplicate-basename-check' option to disable the check of duplicate basenames

Introduce '--no-duplicate-basename-check' option to disable the check of duplicate basenames

With this CL, the check of duplicate basenames in the same source list can be disabled with '--no-duplicate-basename-check' option.

Now GYP generators can handle duplicate basenames except for VCProj generator for Visual C++ 2008 and Makefile generator on Mac. Given that these two generators are no longer actively used, and supposed to be deprecated in future, providing an option to disable this validation should be helpful for some GYP users.

Note that these two generators continue to treat duplicate basenames continue as an error regardless of '--no-duplicate-basename-check'.

BUG=gyp:264, gyp:384
R=scottmg@chromium.org

Committed: https://code.google.com/p/gyp/source/detail?r=1947

[yukawa, 2014-06-18 17:31:02]

Hi Scott, Nico. Torne, Ryan.

I know that I'm not the first person who requested this, but I'd like to hear
comments about the possibility to relax the condition when duplicate basenames
are treated as an error.

My assumption is that relaxing this condition shouldn't be so dangerous as long
as duplicate basenames continue to be detected as an error on platforms where
they are actually troublesome.  What do you think about this approach?

Just FYI,  here is the list of related discussions I found.  Sorry if I missed
some important discussions.
- https://code.google.com/p/gyp/issues/detail?id=264
- https://code.google.com/p/gyp/issues/detail?id=384
- https://groups.google.com/d/msg/gyp-developer/QN-nYWTMUkM/utHOh9kjx9oJ
- https://codereview.chromium.org/12063003/
- https://codereview.chromium.org/12381003/

[scottmg, 2014-06-18 19:47:37]

Per Ryan's comment on the thread linked to, I think this is contrary to gyp
design, though it's also kind of a trivial thing to have this many people fiddle
with at this point.

I assume if make on other platforms can work, then Mac could be made to work
too? I think (?) we can deprecate 2008 now (if v8 has transitioned to something
newer), so then we would be able to keep the lowest-common-denominator
behaviour.

[scottmg, 2014-06-18 19:49:20]

On 2014/06/18 19:47:37, scottmg wrote:
> Per Ryan's comment on the thread linked to, I think this is contrary to gyp
> design, though it's also kind of a trivial thing to have this many people
fiddle
> with at this point.
> 
> I assume if make on other platforms can work, then Mac could be made to work
> too? I think (?) we can deprecate 2008 now (if v8 has transitioned to
something
> newer), so then we would be able to keep the lowest-common-denominator
> behaviour.

Oh, I see now it's a libtool thing. Hm, so I guess we'd have to deprecate both
mac/make and vs2008. I don't know if anyone cares about mac/make or not.

[yukawa, 2014-06-19 00:36:38]

On 2014/06/18 19:49:20, scottmg wrote:
> On 2014/06/18 19:47:37, scottmg wrote:
> > Per Ryan's comment on the thread linked to, I think this is contrary to gyp
> > design, though it's also kind of a trivial thing to have this many people
> fiddle
> > with at this point.
> > 
> > I assume if make on other platforms can work, then Mac could be made to work
> > too? I think (?) we can deprecate 2008 now (if v8 has transitioned to
> something
> > newer), so then we would be able to keep the lowest-common-denominator
> > behaviour.
> 
> Oh, I see now it's a libtool thing. Hm, so I guess we'd have to deprecate both
> mac/make and vs2008. I don't know if anyone cares about mac/make or not.

Thank you for the comment.  Yes, I agree that deprecating these two generators
first is the best and reasonable way.  BTW, deprecating something in a
(somewhat) widely used OSS project is difficult in general.  So I'd like to know
what is the best way to make this happen step by step.

Do you think it is acceptable to introduce a new command line flag like
  --no-duplicate-basenames-check [default:false]
like originally proposed in https://codereview.chromium.org/12063003/ ?

Or, do you think we can now introduce a new command line flag like
  --check-compatibility-with-legacy-generators [default:false]
as an first step toward deprecating vs2008/mac-make generator?

Either is fine with me.  I'd like to choose the first option if it allows us to
reach consensus more easily.  On the other hand, if people agree to have the
second option, it would be a good progress to deprecate these generators.

Any thoughts?

[yukawa, 2014-06-20 18:36:42]

On 2014/06/19 00:36:38, yukawa wrote:
> On 2014/06/18 19:49:20, scottmg wrote:
> > On 2014/06/18 19:47:37, scottmg wrote:
> > > Per Ryan's comment on the thread linked to, I think this is contrary to
gyp
> > > design, though it's also kind of a trivial thing to have this many people
> > fiddle
> > > with at this point.
> > > 
> > > I assume if make on other platforms can work, then Mac could be made to
work
> > > too? I think (?) we can deprecate 2008 now (if v8 has transitioned to
> > something
> > > newer), so then we would be able to keep the lowest-common-denominator
> > > behaviour.
> > 
> > Oh, I see now it's a libtool thing. Hm, so I guess we'd have to deprecate
both
> > mac/make and vs2008. I don't know if anyone cares about mac/make or not.
> 
> Thank you for the comment.  Yes, I agree that deprecating these two generators
> first is the best and reasonable way.  BTW, deprecating something in a
> (somewhat) widely used OSS project is difficult in general.  So I'd like to
know
> what is the best way to make this happen step by step.
> 
> Do you think it is acceptable to introduce a new command line flag like
>   --no-duplicate-basenames-check [default:false]
> like originally proposed in https://codereview.chromium.org/12063003/ ?
> 
> Or, do you think we can now introduce a new command line flag like
>   --check-compatibility-with-legacy-generators [default:false]
> as an first step toward deprecating vs2008/mac-make generator?
> 
> Either is fine with me.  I'd like to choose the first option if it allows us
to
> reach consensus more easily.  On the other hand, if people agree to have the
> second option, it would be a good progress to deprecate these generators.
> 
> Any thoughts?

Thank you for your comments to the discussion thread continued in
https://groups.google.com/d/msg/gyp-developer/DP6ZWsDkW4o/e2IFn8-5lAYJ
My understanding is that there is no objection against at least providing an
option to disable the check of duplicate basenames.

I've updated the CL with --no-duplicate-basename-check. Can you take another
look?

[yukawa, 2014-06-24 16:14:01]

On 2014/06/20 18:36:42, yukawa wrote:
> On 2014/06/19 00:36:38, yukawa wrote:
> > On 2014/06/18 19:49:20, scottmg wrote:
> > > On 2014/06/18 19:47:37, scottmg wrote:
> > > > Per Ryan's comment on the thread linked to, I think this is contrary to
> gyp
> > > > design, though it's also kind of a trivial thing to have this many
people
> > > fiddle
> > > > with at this point.
> > > > 
> > > > I assume if make on other platforms can work, then Mac could be made to
> work
> > > > too? I think (?) we can deprecate 2008 now (if v8 has transitioned to
> > > something
> > > > newer), so then we would be able to keep the lowest-common-denominator
> > > > behaviour.
> > > 
> > > Oh, I see now it's a libtool thing. Hm, so I guess we'd have to deprecate
> both
> > > mac/make and vs2008. I don't know if anyone cares about mac/make or not.
> > 
> > Thank you for the comment.  Yes, I agree that deprecating these two
generators
> > first is the best and reasonable way.  BTW, deprecating something in a
> > (somewhat) widely used OSS project is difficult in general.  So I'd like to
> know
> > what is the best way to make this happen step by step.
> > 
> > Do you think it is acceptable to introduce a new command line flag like
> >   --no-duplicate-basenames-check [default:false]
> > like originally proposed in https://codereview.chromium.org/12063003/ ?
> > 
> > Or, do you think we can now introduce a new command line flag like
> >   --check-compatibility-with-legacy-generators [default:false]
> > as an first step toward deprecating vs2008/mac-make generator?
> > 
> > Either is fine with me.  I'd like to choose the first option if it allows us
> to
> > reach consensus more easily.  On the other hand, if people agree to have the
> > second option, it would be a good progress to deprecate these generators.
> > 
> > Any thoughts?
> 
> Thank you for your comments to the discussion thread continued in
> https://groups.google.com/d/msg/gyp-developer/DP6ZWsDkW4o/e2IFn8-5lAYJ
> My understanding is that there is no objection against at least providing an
> option to disable the check of duplicate basenames.
> 
> I've updated the CL with --no-duplicate-basename-check. Can you take another
> look?

Hi Scott.
Please let me know if you still have any concerns.

Thanks!

[scottmg, 2014-06-24 17:14:31]

My apologies, I missed the update. LGTM

[yukawa, 2014-06-24 23:32:12]

On 2014/06/24 17:14:31, scottmg wrote:
> My apologies, I missed the update. LGTM

Thanks!  Submitting.

[yukawa, 2014-06-25 01:49:03]

Committed patchset #4 manually as r1947 (presubmit successful).