[Gdal-dev] Supported Compiler

[Mateusz Loskot]

Tamas Szekeres wrote:
> Mateusz,
> 
> I'm a bit uncertain what do we mean by declaring the various 
> compilers and platforms 'supported' or 'not supported'.

Tamas,

According to my sense of taste, if we declare a compiler is supported,
then GDAL users can expect their requests for help with building GDAL
using that cmpiler will get higher priority.

Also, they can expect they won't get answer like

"sorry, but your compiler is not supported, so we're not going to take
an steps to fix this problem"

> #1 In my opinion if we declare a compiler as 'supported' we should at
>  least make regular tests  with it. It's not enough that someone has 
> managed to compile the code on that compiler/platform at a given 
> time.

I've assumed that SUNpro and MIPSpro compilers are listed as supported
because there are users who confirm it works. AFAIK, Frank is also
testing it from time to time.

> What happens if a subsequent change works pretty well on some 
> platform but breaks the compilation on another (we can find some 
> examples for this kind of issue in the recent days). It's fairly 
> unconvenient to wait for a bug report from someone who uses that 
> compiler/platform and trying to find which changes caused the 
> problem.

I agree.

> I can eventually say that the supported compilers should be part of
> the gdal buildbot configuration.

Personally, I'd really like to see all supported environments connected,
but it's not easy to find free access to sophisticated platforms.

> #2 We should also have a detailed documentation for the supported 
> compilers, how to make the compilation and the tests (for making sure
>  that the compilation was successful).

If I will get access to SUNpro and MIPSpro, I'm eager to write
this docs.

> [...] Therefore the compilation should be reproduceable on a
> 'supported' compiler/platform.

Agreed.

> #3 We should make further declarations about which components are 
> supported on which compilers/platforms because some components might 
> not compile on the same environment where the core gdal builds. For 
> example, apparently, not all of the SWIG bindings maintainers might 
> be aware of all of the compilers/platforms declared as supported now,
>  he might not have the proper environment to test them all.

I didn't do it intentionally.
Compilation support means "what C/C++ compilator can be used to build
GDAL binaries".

If we want to list all components, then we have to:
- list editions/versions of C# compilators
- --//-- for Java
- versions of Python and Perl interpreters

I believe we should focus on the core binaries.

> #4 We should also identify the substantial compiler configurations 
> that will require significantly different approach when making the 
> builds/tests. In my opinion the difference between the Win32/Win64 
> builds are a good candidate to make this separation. The document 
> describing the Win64 with VC8 would be somewhat different as the 
> description of the Win32 with VC8.

I agree.

> So after all I can say that at first we should try to tuck the 
> supported component/compiler/platform combination into the buildbot 
> configuration and after the successful assimilation we can surely 
> make the declaration that this configuration is supported or not.

Agreed but I have no idea who to ask for platforms
like SUN with SUNpro and SGI with MIPSpro.

Cheers
-- 
Mateusz Loskot
http://mateusz.loskot.net