Re: [Plplot-devel] Question about Ada support

[Tom K. <tom...@ve...>]

On Tue, May 19, 2015 at 7:22 PM, Alan W. Irwin <ir...@be...>
wrote:

> On 2015-05-19 13:20-0400 Tom Kacvinsky wrote:
>
>  Alan,
>>
>> This is not in regards to plplot proper, but rather the cmake Ada support
>> bundled with plplot.
>>
>> How is add_executable supposed to work?  Does it invoke gcc for each
>> adb/ads file, then call the rest of the GNAT tool chain utilities to
>> generate an executable, or does it just invoke gnatmake?  the reason I ask
>> is that we aren't listing all of the Ada files necessary to build the
>> executable, but all of the necessary files end up getting compiled.  But
>> upon a subsequent build, not all Ada files get recompiled, even if they've
>> changed.  I am thinking if gnatmake is used, this change would be detected
>> and the right thing would happen.
>>
>> So, should I be listing all of the necessary Ada files in the
>> add_executable(<target> <files>) invocation so that no matter what is used
>> under the hood, recompiles happen when necessary.
>>
>
> Hi Tom:
>
> From memory I am pretty sure that gnatmake is used under the hood for
> the add_executable case.  But the thing to do is to run
>
> make VERBOSE=1
>
> (or verbose equivalent if you are building with a non-make generator) to
> see exactly what commands are being used for the Ada executable build.
>
> And to answer your last question, I am virtually positive you should
> mention all necessary Ada source files in the add_executable(<target>
> <files>) invocation because that is the way add_excutable is supposed
> to work in general.  But to figure it out for sure try with and
> without mentioning all Ada source files for VERBOSE=1 to see the
> exact difference that makes to the build.
>
> Sorry that I am being a little vague here, but our own use of Ada
> language support only uses single files in the add_executable command
> so I don't know exactly what would happen for multiple files. Furthermore,
> it has been a very long while since I worked on our Ada
> language support, and when I implemented that I had very little
> understanding of CMake language support (and that is still the case).
> So I converted existing language support for C over to Ada language
> support using trial and error methods using tests like I suggested
> above to evaluate whether the results were good enough for our needs.
> And stopped when that was finally the case.
>

Hi again Alan,

I made the changes to add_executable and it appears gnatmake is not invoked
to build the executable, at least not that I can tell with "make
VERBOSE=1".  gcc is getting invoked to make object files from the Ada
source, but the problem I am now running into is that cmake is now using
the system gcc to compile the files, where we want a different gcc to be
used.  IS there a way of telling the cmake Ada support to use a different
gcc (like there is for C, C++ and FORTRAN compilers)?  I am talking about
tricks with cmake variables, not the PATH setting (as then the gcc that we
want for Ada would be used for C/C++, and for those compiles, we want a
different gcc).

Thanks,

Tom