cmake for python projects

[Rex Dieter]

On 09/28/2010 02:39 PM, Alexander Neundorf wrote:
> On Monday 27 September 2010, Wolfgang Rohdewald wrote:
>> On Montag 27 September 2010, Jonathan Riddell wrote:
>>> README.packagers will do but I still prefer my method of
>>> giving a build time error unless the packager has overridden
>>> it.
>>>
>>> "Developers do not need cmake for Python development at all."
>>> this isn't necessarily right, there might well be items that
>>> need to be installed in the right place.
>>
>> I deleted that.
>>
>>> "Distribution packagers do NOT want cmake to error out on
>>> missing runtime dependencies" as above, I'm a packager and
>>> I'd rather it did, I'd certainly prefer to have warnings.
>>
>> I changed that to "Most distribution packagers" (namely
>> Sune Vuorela/Debian and Johannes Obermayr/Suse do not want
>> them, while you do. I do not know about others).
> 
> I think you can also count Dirk Mueller in.
> I have put Rex Dieter from Fedora/RedHat on CC, so he can also say how he sees 
> the issue.
> 
> I think we need to decide for *one policy*.
> 
> While I as a developer like it if I get build/cmake time errors for runtime 
> dependencies (so I can fix them and know when the thing is installed, that 
> everything will be there and it will work), the distro packagers basically 
> convinced that the right thing to do is really only to depend on thing at 
> build/cmake time which are actually required at build/cmake time.
> 
> I.e., no runtime dependency checking.

+1 here too, no fatal run-time dep checks at build-time.

-- Rex