[cmake-developers] FindX11 change breaks complex test

[David Cole]

On Wed, Aug 15, 2012 at 3:31 PM, Rolf Eike Beer <eike at sf-mail.de> wrote:

> David Cole wrote:
> > On Wed, Aug 15, 2012 at 1:28 PM, Rolf Eike Beer <eike at sf-mail.de> wrote:
> > > Bill Hoffman wrote:
> > >
> > >
> http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=97c338bac910d087eff9b16
> > > 0e>
> > > > 3b68f2eeee482a5
> > > >
> > > >
> > > > http://open.cdash.org/testDetails.php?test=156453780&build=2517168
> > > >
> > > > CMake Error at /export/home/kitware/Dashboards/My
> > > >
> > > > Tests-SunOSsnv_135.11i86pc-gcc/CMake/Modules/FindX11.cmake:329
> (list):
> > > >    list sub-command REMOVE_DUPLICATES requires list to be present.
> > > >
> > > > Call Stack (most recent call first):
> > > >    /export/home/kitware/Dashboards/My
> > >
> > >
> Tests-SunOSsnv_135.11i86pc-gcc/CMake/Modules/CMakeBackwardCompatibilityC.c
> > > ma>
> > > > ke:56 (include)
> > > >
> > > >    CMakeLists.txt:5 (project)
> > >
> > > I'll work around this for now, but I wonder if the restriction in
> > > list(REMOVE_DUPLICATES) actually makes sense. We don't give an error if
> > > there
> > > are no duplicates, why should we give an error if there are no items?
> >
> > Is it that there are no items, or that there is not even a list with that
> > name?
>
> In line 151 of FindX11.cmake there is a
>
>   set(X11_LIBRARY_DIR "")
>
> so the variable should exists.
>
> Eike
> --
> --
>
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the CMake FAQ at:
> http://www.cmake.org/Wiki/CMake_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers
>


Well, I can't think of why anybody would care if REMOVE_DUPLICATES on an
empty list produces a still empty list with no error.

I suppose somebody could have some code that relies on the error
occurring... Perhaps the error is useful to some as a "possible typo in
variable name" indicator.

But really, we should probably just be silent if there *is* a variable of
that name, and it has no items.

It would be a behavior change -- how strict should we be for backwards
compatibility in this case?


:-)  -- in the metaphysical sense, this gets into the question of what is a
list... is a CMake variable value with exactly 0 or 1 "items" in it
actually a list...? Or is it not a "list" until there are 2 or more items
in it.


Silliness aside, I would be ok with changing the behavior to avoid the
error in this particular case. Unless somebody has a real-world
counter-argument where it is considered important to preserve our presently
silly behavior.


Thx,
David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/cmake-developers/attachments/20120815/b5b998c7/attachment.html>