llvm dev - Mar 2015 - [LLVMdev] [RFC] Raise minimum required CMake version to 3.0

> We don't require users to compile their own gcc. Are you saying
it's
> impossible to get pre-built CMake 2.8.12 and gcc 4.7 on either Mac or
> Windows?
And we wouldn't be requiring it for cmake. My point was that cmake is
always easier to install:

* From binaries: a bit easier, since they are provided in http://www.cmake.org/
* From source: way easier than gcc.

Gosh, from my experience *building* cmake is easier than *downloading*
llvm and clang in the right layout, so the extra cost is really small.

As for the advantages, this seems to make it easier to drop the
autoconf build, which would be a really big win for us.
>> We already require it of every developer using windows or os X. By the
>> same argument we should be providing xcode and msvc projects :-)
>
> Why not? CMake can build eclipse files, and some projects do supply
> code::blocks project files. If people feel inclined, I don't see a
problem.
Because it has a big cost for the people actually writing llvm in
exchange of an hypothetical advantage to an hypothetical user.

Cheers,
Rafael

On 3/10/15 6:46 PM, Rafael Espíndola wrote:
>> We don't require users to compile their own gcc. Are you saying
it's
>> impossible to get pre-built CMake 2.8.12 and gcc 4.7 on either Mac or
>> Windows?
>
> And we wouldn't be requiring it for cmake. My point was that cmake is
> always easier to install:
>
> * From binaries: a bit easier, since they are provided in
http://www.cmake.org/
> * From source: way easier than gcc.
>
> Gosh, from my experience *building* cmake is easier than *downloading*
> llvm and clang in the right layout, so the extra cost is really small.
>
> As for the advantages, this seems to make it easier to drop the
> autoconf build, which would be a really big win for us.
>
>>> We already require it of every developer using windows or os X. By
the
>>> same argument we should be providing xcode and msvc projects :-)
>>
>> Why not? CMake can build eclipse files, and some projects do supply
>> code::blocks project files. If people feel inclined, I don't see a
problem.
>
> Because it has a big cost for the people actually writing llvm in
> exchange of an hypothetical advantage to an hypothetical user.
Enthusiastic +1
>
> Cheers,
> Rafael
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
-- 
Jon Roelofs
jonathan at codesourcery.com
CodeSourcery / Mentor Embedded

On Tue, Mar 10, 2015 at 5:46 PM, Rafael Espíndola <
rafael.espindola at gmail.com> wrote:
> As for the advantages, this seems to make it easier to drop the
> autoconf build, which would be a really big win for us.
>
My only problem here is this: it *seems*.

I would like concrete and specific advantages. I think we're just being way
too hypothetical and vague.

If there are specific things that we cannot do today and could do by
requiring a certain version of cmake, that would be a good discussion to
have. Saying that there might be things and that they might help isn't
going to get us anywhere. ;]
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150310/0cf1daed/attachment.html>

> > We don't require users to compile their own gcc. Are you saying
it's
> > impossible to get pre-built CMake 2.8.12 and gcc 4.7 on either Mac or
> > Windows?
> 
> And we wouldn't be requiring it for cmake. My point was that cmake is
> always easier to install:
> 
> * From binaries: a bit easier, since they are provided in
> http://www.cmake.org/
> * From source: way easier than gcc.
> 
> Gosh, from my experience *building* cmake is easier than *downloading*
> llvm and clang in the right layout, so the extra cost is really small.
Enough "really small" costs add up to too much.  Not so long ago it
took
me a couple days to get the right versions of enough packages sorted out
(on Linux) so that I could get the thing I actually wanted to work.  I
had to want it pretty badly to get through that. I wouldn't wish that on
anybody.  Probably makes me over-sensitive to prerequisite costs but...
not something to require lightly.
--paulr

On 10 March 2015 at 21:08, Chandler Carruth <chandlerc at google.com>
wrote:
>
> On Tue, Mar 10, 2015 at 5:46 PM, Rafael Espíndola
> <rafael.espindola at gmail.com> wrote:
>>
>> As for the advantages, this seems to make it easier to drop the
>> autoconf build, which would be a really big win for us.
>
>
> My only problem here is this: it *seems*.
>
> I would like concrete and specific advantages. I think we're just being
way
> too hypothetical and vague.
>
> If there are specific things that we cannot do today and could do by
> requiring a certain version of cmake, that would be a good discussion to
> have. Saying that there might be things and that they might help isn't
going
> to get us anywhere. ;]
>From above in this thread:
--------------------------------------------------
I should also point out that CMAKE_SYSROOT and
CMAKE_<LANG>_COMPILER_TARGET (both CMake 3.0 features) would make
fixing compiler-rt's CMake (Bugs 14109 & 21562) a lot easier. Both of
those bugs are currently blockers to depreciating the autotools build
system.
-------------------------------------------------

But it does look like we have a general issue here: Why is linux
special? If requiring a cmake that does't ship with a given system is
an issue at all, then we couldn't use any version of cmake because of
OS X and Windows.

Cheers,
Rafael

> On Mar 10, 2015, at 6:08 PM, Chandler Carruth <chandlerc at
google.com> wrote:
> 
> 
> On Tue, Mar 10, 2015 at 5:46 PM, Rafael Espíndola <rafael.espindola at
gmail.com <mailto:rafael.espindola at gmail.com>> wrote:
> As for the advantages, this seems to make it easier to drop the
> autoconf build, which would be a really big win for us.
> 
> My only problem here is this: it *seems*.
> 
> I would like concrete and specific advantages. I think we're just being
way too hypothetical and vague.
> 
> If there are specific things that we cannot do today and could do by
requiring a certain version of cmake, that would be a good discussion to have.
Saying that there might be things and that they might help isn't going to
get us anywhere. ;]
Chris provided a detailed list of desirable features for us in newer versions of
CMake in this very thread:

http://permalink.gmane.org/gmane.comp.compilers.llvm.devel/82824
<http://permalink.gmane.org/gmane.comp.compilers.llvm.devel/82824>

And further elaborated on it:

http://permalink.gmane.org/gmane.comp.compilers.llvm.devel/82793
<http://permalink.gmane.org/gmane.comp.compilers.llvm.devel/82793>
http://permalink.gmane.org/gmane.comp.compilers.llvm.devel/83315
<http://permalink.gmane.org/gmane.comp.compilers.llvm.devel/83315>

I’m not sure what further degree of specificity you’re looking for.

—Owen
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150310/1e2f9289/attachment.html>