build-system for glas

[Toon Knapen]

I'm wondering what build-system we should use for building the tests of
glas.

Currently we are using bjam. Bjam is multi-platform and has integrated
support for regression testing. Drawback is that bjam has a rather steep
learning curve.

An other option would be to use makefile's (on posix) and visual studio
solution files on windows. This means however that we will have to sync
2 build systems and that we have no integrated support for regression
testing.

Opinions?

t
_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

Toon Knapen wrote:
> Opinions?

The KDE project recently switched to CMake (http://www.cmake.org). It
was a huge success. CMake is cross-platform, easy to learn and use, and
has proven powerful enough to handle one of the biggest open-source
projects around.
_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

On Thursday 04 January 2007 10:33 am, Simon Perreault wrote:
> Toon Knapen wrote:
> > Opinions?
>
> The KDE project recently switched to CMake (http://www.cmake.org). It
> was a huge success. CMake is cross-platform, easy to learn and use, and
> has proven powerful enough to handle one of the biggest open-source
> projects around.

I have been quite happy with scons.
_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Georg Baum]

Am Donnerstag, 4. Januar 2007 16:09 schrieb Toon Knapen:

> I'm wondering what build-system we should use for building the tests of
> glas.
>
> Currently we are using bjam. Bjam is multi-platform and has integrated
> support for regression testing. Drawback is that bjam has a rather steep
> learning curve.
>
> An other option would be to use makefile's (on posix) and visual studio
> solution files on windows. This means however that we will have to sync
> 2 build systems and that we have no integrated support for regression
> testing.
>
> Opinions?

Syncing two build systems is only asking for trouble. Both scons and cmake
seem to be suitable, but the learning curve is still considerable IMHO. If
the learning curve is the only reason for switching I would probably not
do it, but rather document the jam files well.


Georg

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Doug Gregor]

On Fri, 2007-01-05 at 11:52 +0100, Georg Baum wrote:

> Am Donnerstag, 4. Januar 2007 16:09 schrieb Toon Knapen:
> > I'm wondering what build-system we should use for building the tests of
> > glas.
> >
> > Currently we are using bjam. Bjam is multi-platform and has integrated
> > support for regression testing. Drawback is that bjam has a rather steep
> > learning curve.
> >
> > An other option would be to use makefile's (on posix) and visual studio
> > solution files on windows. This means however that we will have to sync
> > 2 build systems and that we have no integrated support for regression
> > testing.
> >
> > Opinions?
>
> Syncing two build systems is only asking for trouble. Both scons and cmake
> seem to be suitable, but the learning curve is still considerable IMHO. If
> the learning curve is the only reason for switching I would probably not
> do it, but rather document the jam files well.

With "bjam", are you currently using Boost Build version 1? I ask
because with the next release, Boost is moving to Boost Build version 2,
which is very, very different from version 1... almost to the point of
being a completely new build system.

So, the learning curve might remain even if you keep using bjam.

  Cheers,
  Doug

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Karl Meerbergen-2]

Douglas Gregor wrote:

>With "bjam", are you currently using Boost Build version 1? I ask
>because with the next release, Boost is moving to Boost Build version 2,
>which is very, very different from version 1... almost to the point of
>being a completely new build system.
>
>So, the learning curve might remain even if you keep using bjam.
>
>  Cheers,
>  Doug
>
>  
>

Doug,

We're using v2

Karl

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

[Karl.Meerbergen.vcf]

---

begin:vcard
fn:Karl Meerbergen
n:Meerbergen;Karl
org:Katholieke Universiteit Leuven;Department of Computer Science
adr:;;Celestijnenlaan 200A;Heverlee (Leuven);;B-3001;Belgium
email;internet:Karl.Meerbergen@...
tel;work:[phone]
tel;fax:[phone]
url:http:://www.cs.kuleuven.be/~karlm
version:2.1
end:vcard

---

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Philipp Schwaha]

might I suggest you take a look at waf:
http://freehackers.org/~tnagy/bksys.html

greetings
        Philipp

On Thu, 2007-01-04 at 11:18 -0500, Neal Becker wrote:

> On Thursday 04 January 2007 10:33 am, Simon Perreault wrote:
> > Toon Knapen wrote:
> > > Opinions?
> >
> > The KDE project recently switched to CMake (http://www.cmake.org). It
> > was a huge success. CMake is cross-platform, easy to learn and use, and
> > has proven powerful enough to handle one of the biggest open-source
> > projects around.
>
> I have been quite happy with scons.
> _______________________________________________
> glas mailing list
> glas@...
> http://lists.boost.org/mailman/listinfo.cgi/glas
>

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Karl Meerbergen-2]

Hi,

Waf is python based. What are the advantages compared to scons?

Whatever the choice of build system is, I think it is important that
fortran should also be able to be compiled. I recall that we (at FFT)
had some problems with making bindings to blas 1 functions on some
platforms. The only way to circonvent this was making a subroutine in
fortran and using this in the blas backend. Perhaps I should first check
whether this is still the case and identify more precisely what the
problem was.

However, another point is that we will also perform performance tests on
various platforms. Ideally these will be performed with the same build
system as the regression tests. I would like to make comparisons with
some fortran codes, not just other c++ or c codes.

Karl



Philipp Schwaha wrote:

>might I suggest you take a look at waf:
>http://freehackers.org/~tnagy/bksys.html
>
>greetings
> Philipp
>
>On Thu, 2007-01-04 at 11:18 -0500, Neal Becker wrote:
>  
>
>>On Thursday 04 January 2007 10:33 am, Simon Perreault wrote:
>>    
>>
>>>Toon Knapen wrote:
>>>      
>>>
>>>>Opinions?
>>>>        
>>>>
>>>The KDE project recently switched to CMake (http://www.cmake.org). It
>>>was a huge success. CMake is cross-platform, easy to learn and use, and
>>>has proven powerful enough to handle one of the biggest open-source
>>>projects around.
>>>      
>>>
>>I have been quite happy with scons.
>>_______________________________________________
>>glas mailing list
>>glas@...
>>http://lists.boost.org/mailman/listinfo.cgi/glas
>>
>>    
>>
>
>_______________________________________________
>glas mailing list
>glas@...
>http://lists.boost.org/mailman/listinfo.cgi/glas
>  
>

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

[Karl.Meerbergen.vcf]

---

begin:vcard
fn:Karl Meerbergen
n:Meerbergen;Karl
org:Katholieke Universiteit Leuven;Department of Computer Science
adr:;;Celestijnenlaan 200A;Heverlee (Leuven);;B-3001;Belgium
email;internet:Karl.Meerbergen@...
tel;work:[phone]
tel;fax:[phone]
url:http:://www.cs.kuleuven.be/~karlm
version:2.1
end:vcard

---

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Toon Knapen]

I mainly hesitated about the build-system because I was not sure that we
should use a more 'advanced' build-system as plain makefiles with the
disadvantage of making the learning curve a bit higher.

All suggestions were proposing one of these more 'advanced'
build-systems. All of these (bjam, scons, ...) have a learning curve so
whichever we pick, users will need time to get used to the build system

So, since we are already using boost a lot in glas, I therefor think we
best continue using bjam so that at least people that are involved in
both projects only have to deal with one build system.

t

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Philipp Schwaha]

waf is much faster than scons and also supports unit tests, which can be
done with scons, but are in my experience quite cumbersome.

I have not tried any fortran code yet and have not found any examples
for it, I will have to look into this!

greetings
        Philipp

On Thu, 2007-01-11 at 14:36 +0100, Karl Meerbergen wrote:

> Hi,
>
> Waf is python based. What are the advantages compared to scons?
>
> Whatever the choice of build system is, I think it is important that
> fortran should also be able to be compiled. I recall that we (at FFT)
> had some problems with making bindings to blas 1 functions on some
> platforms. The only way to circonvent this was making a subroutine in
> fortran and using this in the blas backend. Perhaps I should first check
> whether this is still the case and identify more precisely what the
> problem was.
>
> However, another point is that we will also perform performance tests on
> various platforms. Ideally these will be performed with the same build
> system as the regression tests. I would like to make comparisons with
> some fortran codes, not just other c++ or c codes.
>
> Karl
>
>
>
> Philipp Schwaha wrote:
>
> >might I suggest you take a look at waf:
> >http://freehackers.org/~tnagy/bksys.html
> >
> >greetings
> > Philipp
> >
> >On Thu, 2007-01-04 at 11:18 -0500, Neal Becker wrote:
> >  
> >
> >>On Thursday 04 January 2007 10:33 am, Simon Perreault wrote:
> >>    
> >>
> >>>Toon Knapen wrote:
> >>>      
> >>>
> >>>>Opinions?
> >>>>        
> >>>>
> >>>The KDE project recently switched to CMake (http://www.cmake.org). It
> >>>was a huge success. CMake is cross-platform, easy to learn and use, and
> >>>has proven powerful enough to handle one of the biggest open-source
> >>>projects around.
> >>>      
> >>>
> >>I have been quite happy with scons.
> >>_______________________________________________
> >>glas mailing list
> >>glas@...
> >>http://lists.boost.org/mailman/listinfo.cgi/glas
> >>
> >>    
> >>
> >
> >_______________________________________________
> >glas mailing list
> >glas@...
> >http://lists.boost.org/mailman/listinfo.cgi/glas
> >  
> >
>
>
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
> _______________________________________________
> glas mailing list
> glas@...
> http://lists.boost.org/mailman/listinfo.cgi/glas

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

scons certainly supports F***TRAN
_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas

[Karl Meerbergen-2]

For me, bjam is ok. I know a little, enough to add new regression tests,
but not enough to set up a whole build system. If there is no strong
argument to change, I would stick to bjam.

Karl



Neal Becker wrote:

>scons certainly supports F***TRAN
>_______________________________________________
>glas mailing list
>glas@...
>http://lists.boost.org/mailman/listinfo.cgi/glas
>  
>


Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

[Karl.Meerbergen.vcf]

---

begin:vcard
fn:Karl Meerbergen
n:Meerbergen;Karl
org:Katholieke Universiteit Leuven;Department of Computer Science
adr:;;Celestijnenlaan 200A;Heverlee (Leuven);;B-3001;Belgium
email;internet:Karl.Meerbergen@...
tel;work:[phone]
tel;fax:[phone]
url:http:://www.cs.kuleuven.be/~karlm
version:2.1
end:vcard

---

_______________________________________________
glas mailing list
glas@...
http://lists.boost.org/mailman/listinfo.cgi/glas