[Qt5-feedback] Build system requirements for Qt5

[Oswald Buddenhagen]

On 06/07/11 03:09, ext Bill Hoffman wrote:
> On 6/6/2011 7:12 PM, Thiago Macieira wrote:
>> Em Tuesday, 7 de June de 2011, às 00:52:16, Alexander Neundorf escreveu:
>>> srcs = foo.c, bar.c, blub.c
>>>
>>> if(Windows)
>>>      srcs += win.c
>>>
>>> if (HAVE_MOUNT_H)
>>>      srcs += mount.c
>>>
>>> if(BSD || HAVE_WHATEVER)
>>>      if (SOME_VERSION<   1.0)
>>>         srcs += x1.c
>>>      else
>>>         srcs += x2.c
>>>
>>> add_executable(hello $srcs)
>>>
>>> So, I'm curious how you plan to solve this.
>>
>> You're trying to solve the problem of having a GUI for writing every aspect of
>> the build system. That's not what we want.
>>
while the general sentiment is true, this does certainly not apply to 
the above example, which is rather common code. it would be ok to forbid 
nesting, so the scope conditionals could be presented as a flat list, 
but that's about it.

>> What we want from the above is the full listing of source files (to present in
>> the project listing).
>>
>
> Then, you would have some other cmake file that used that list and took
> out the files that were not used on the current system.
>
> if(NOT WIN32)
>     remove_sources(sources win.c)
>
this is insane. ide integration must not come at the cost of manual 
usability.

> Certainly, before any new
> system gets started,  it would be nice to spec out the functionality
>
we did. after implementing it, cmake wouldn't be cmake any more. so why 
bother?
but there is certainly room for borrowing some code or at least 
algorithms - cmComputeLinkInformation.cxx for example is a bonanza of 
well-researched obscure stuff.