Closed GoogleCodeExporter closed 9 years ago
Any plans for this? It is blocking one of my CLs.
Original comment by mtomasz@chromium.org
on 27 Mar 2014 at 3:40
I think we could probably delete the warning at this point, but there are some
non-Chromium users of gyp that are still using 2008. Presumably you could just
rename one of your files as we've been doing up until now?
Original comment by scottmg@chromium.org
on 27 Mar 2014 at 5:11
I've been applying brandon...@gmail.com's patch which is attached to issue
https://code.google.com/p/gyp/issues/detail?id=264
that downgrades the error to a warning if the generator is not msvs2008.
It has applied cleanly to all of the gyp repos revisions I've updated to,
including very recent ones. I recommend that it be reviewed and integrated.
/Steve Hartwell
https://gyp.googlecode.com/issues/attachment?aid=2640002000&name=0b2eb1f758bbe24
75626e8a1a8e4fcebaeb027a9.patch.txt&token=jXE76LOB3jiPqA6vUuRJVpVqH_w%3A13959836
26090
Original comment by steve.ha...@gmail.com
on 28 Mar 2014 at 5:22
@scottmg: Whith the amount of files in Chromium it is sometimes very difficult
to rename. If you use namespaces well to avoid long class names, then this
becomes a problem.
Eg. I wanted to have a AsyncFileUtil class in my own file_system_provider
namespace to avoid FileSystemProviderAsyncFileUtil longish class name. But it
can't be done because of this GYP error. As a result, I had to add some
unnatural prefix like ProviderAsyncFileUtil, which is inconsistent in naming
with all other classes in the namespace.
How about adding some switch enabling support for vs2008, which would be
disabled by default?
Original comment by mtomasz@chromium.org
on 28 Mar 2014 at 5:30
FYI, in gyp r1947 and later, you can opt out this validation with
'--no-duplicate-basename-check' GYP option.
See the following threads about the discussion behind r1947.
https://groups.google.com/d/msg/gyp-developer/DP6ZWsDkW4o/e2IFn8-5lAYJ
https://groups.google.com/d/msg/gyp-developer/qQ7RxD7yCUg/GRn7AbUtygAJ
Hopefully this validation will be retired gradually (opt-in -> opt-out ->
removed), but I'm not sure when it will happen though.
As an interim solution for Chromium, it might be acceptable to add
'--no-duplicate-basename-check' to src/build/gyp_chromium because I guess we no
longer care about VC++2008 and make on Mac in Chromium. This could be
justified because we have already relied on '--no-circular-check' for Mac in
gyp_chromium.
https://code.google.com/p/chromium/codesearch#chromium/src/build/gyp_chromium
Once Chromium have accepted '--no-duplicate-basename-check', we can safely
remove the following workaround from GYP. It should be helpful for GYP.
https://code.google.com/p/gyp/source/browse/trunk/pylib/gyp/input.py?r=1947#2842
Perhaps it should also be helpful for GN folks because I guess we don't want GN
to inherit this limitation from GYP.
Original comment by yukawa@chromium.org
on 25 Jun 2014 at 2:48
Fixed as r1993 by a patch from Matthew.
Original comment by yukawa@chromium.org
on 24 Oct 2014 at 1:53
Turns out this isn't obsolete after all, libtool prints a warning about this on
OS X.
Original comment by thakis@chromium.org
on 21 May 2015 at 7:38
(See e.g. "[chromium-dev] Does chromium require .cc/.h files to have unique
basenames?" – this should've been a gyp-time thing, not a
land-revert-since-it-added-a-warning)
Original comment by thakis@chromium.org
on 21 May 2015 at 7:39
Sorry for the trouble. When I wrote r1947, I confirmed that max/xcode can
handle duplicated base names (by internally renaming them) in a static library,
but apparently I should have checked and mentioned in the case of mac/ninja
scenario.
Original comment by yukawa@chromium.org
on 21 May 2015 at 8:12
Original issue reported on code.google.com by
scottmg@chromium.org
on 23 Nov 2013 at 3:42