[cmake-developers] Questions on staged topic branches

[David Cole]

On Tue, Sep 11, 2012 at 6:06 PM, James Bigler <jamesbigler at gmail.com> wrote:
> On Tue, Sep 4, 2012 at 11:48 AM, David Cole <david.cole at kitware.com> wrote:
>>
>> If you are on the "To" line, please reply to this email, and answer
>> the questions below:
>>
>> Brad and I have questions on these staged topic branches before we
>> consider them for merging to 'master'.
>>
>>
>> ##########
>> # James Bigler:
>> #             topics/FindCUDA/Add-CUDA_HOST_COMPILER | master=0 next=1
>>
>> Is CUDA_HOST_COMPILER a path to a directory or a file? It appears to
>> take on different values depending on which if branch you go through.
>> Is that true? Or should it always be a full path to the compiler?
>>
>
> It can technically be either.
>
> --compiler-bindir <path>                    (-ccbin)
>         Specify the directory in which the compiler executable (Microsoft
>         Visual Studio cl, or a gcc derivative) resides. By default, this
>         executable is expected in the current executable search path. For a
>         different compiler, or to specify these compilers with a different
>         executable name, specify the path to the compiler including the
>         executable name.
>
> What is the best course of action in this case?  What I've implemented seems
> to work, so I didn't think too much about it.
>
> As far as rebasing on the changes to the compiler id stuff, I would rather
> not until it becomes a must have feature.  I've tried to make FindCUDA work
> on older than TOT versions of CMake to facilitate my local developer pool.
>
> James
>

I was just asking the questions for clarification. If you think it is
best as-is, and do not want to make further changes to it, we can
accept it as-is into master. I think a variable that could be either a
full path name to a compiler, or just the directory that contains a
compiler (if it's this one, how do you know what the compiler in the
directory is...?) -- I think it's just confusing and unclear.

But if it works as expected for people who are using FindCUDA, I guess
it's ok. (as you can probably tell, I'm not one of them...)

We'll merge it next time around, unless you reply again and say you
want to make further changes on the topic first.


Thanks for the reply,
David