www.digitalmars.com         C & C++   DMDScript  

c++ - Boost configuration for DM?

reply John Fletcher <J.P.Fletcher aston.ac.uk> writes:
Has anyone done the Boost configuration for the DM compiler?

Thanks

John
Oct 15 2002
parent reply John Fletcher <J.P.Fletcher aston.ac.uk> writes:
John Fletcher wrote:


 Has anyone done the Boost configuration for the DM compiler?

 Thanks

 John


I am setting about doing this from scratch, and have the structure set
up, but has anyone ever run configure test for Boost and could tell me
how to do it?  I cannot find anything on the Boost sites about how to do
this.

John Fletcher
Oct 16 2002
parent reply "Walter" <walter digitalmars.com> writes:
I haven't looked at the boost stuff. Christof has, and he says that DM's
namespace support has to get better before it will work. -Walter

"John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
news:3DAD5032.8C99099B aston.ac.uk...

 John Fletcher wrote:


 Has anyone done the Boost configuration for the DM compiler?

 Thanks

 John


 I am setting about doing this from scratch, and have the structure set
 up, but has anyone ever run configure test for Boost and could tell me
 how to do it?  I cannot find anything on the Boost sites about how to do
 this.

 John Fletcher
Oct 16 2002
parent reply "Matthew Wilson" <dmd synesis.com.au> writes:
I would suggest that would be the case also. The last time I looked, they
don't have a facility for disabling the boost namespace (which stlsoft does
have, ;)), so any out-of-class template-class member functions are likely to
give nasties

"Walter" <walter digitalmars.com> wrote in message
news:aokdr2$8tb$1 digitaldaemon.com...

 I haven't looked at the boost stuff. Christof has, and he says that DM's
 namespace support has to get better before it will work. -Walter

 "John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
 news:3DAD5032.8C99099B aston.ac.uk...

 John Fletcher wrote:


 Has anyone done the Boost configuration for the DM compiler?

 Thanks

 John


 I am setting about doing this from scratch, and have the structure set
 up, but has anyone ever run configure test for Boost and could tell me
 how to do it?  I cannot find anything on the Boost sites about how to do
 this.

 John Fletcher
Oct 16 2002
next sibling parent reply John Fletcher <J.P.Fletcher aston.ac.uk> writes:
Matthew Wilson wrote:


 I would suggest that would be the case also. The last time I looked, they
 don't have a facility for disabling the boost namespace (which stlsoft does
 have, ;)), so any out-of-class template-class member functions are likely to
 give nasties

 "Walter" <walter digitalmars.com> wrote in message
 news:aokdr2$8tb$1 digitaldaemon.com...

 I haven't looked at the boost stuff. Christof has, and he says that DM's
 namespace support has to get better before it will work. -Walter

 "John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
 news:3DAD5032.8C99099B aston.ac.uk...

 John Fletcher wrote:


 Has anyone done the Boost configuration for the DM compiler?

 Thanks

 John


 I am setting about doing this from scratch, and have the structure set
 up, but has anyone ever run configure test for Boost and could tell me
 how to do it?  I cannot find anything on the Boost sites about how to do
 this.

 John Fletcher






I note the assessments above.  I spent some time on this yesterday and can
report the following results.

First of all, how to configure boost.  The advice in the boost community is
basically

1.  run ./configure on Unix.
2.  read the documents.  These do not fully correspond with the actual test
programs.

so I am back to trial and error.

There is a file called boost\config\select_compiler_config.hpp and I have added
to it the following section

#elif defined __DMC__
//  Digital Mars put in BEFORE MPW as DM also defines __SC__ (Symantec)
//  for backward compatibility but I have NOT tested that.


which sets up dmc.hpp for Digital Mars.  Note that there is another compiler
called MPW which also defines __SC__ so it is important to get in before that
one.

After that the question is what to put into dmc.hpp.

There are two programs to help with this.

boost/libs/config/test/config_info.cpp
boost/libs/config/test/config_test.cpp

Status:

config_info.cpp will not compile with DM8.30 owing to lack of <iostream>
It will compile and run with DM8.30 +STLport and provides some useful
configuration information.

config_test.cpp will not compile with DM8.30 + STLport
or with DM8.31.1 + STLport and the compilation error does not change.

I am currently investigating this to pin it down.  It could be solveable by a
BOOST config choice (although I think I have tried most of those already) or it
could need a change to the DM compiler.

If anyone else is interested I will share the files mentioned above.

John Fletcher
Oct 17 2002
next sibling parent "Matthew Wilson" <dmd synesis.com.au> writes:
John

I'm interested, but am about to take a holiday, so may not be able to help
for a couple of weeks. Upon my return, Boost, WinSTL and Comeau are my top
priorities, all of which involve tinkering with compilers and stuff, so if
you are still in need of assistance by then I'll definitely be up for it.

Please send me what you have now, as I am interested in taking a look

Matthew


"John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
news:3DAE7694.C9958229 aston.ac.uk...

 Matthew Wilson wrote:


 I would suggest that would be the case also. The last time I looked,




they

 don't have a facility for disabling the boost namespace (which stlsoft




does

 have, ;)), so any out-of-class template-class member functions are




likely to

 give nasties

 "Walter" <walter digitalmars.com> wrote in message
 news:aokdr2$8tb$1 digitaldaemon.com...

 I haven't looked at the boost stuff. Christof has, and he says that






DM's

 namespace support has to get better before it will work. -Walter

 "John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
 news:3DAD5032.8C99099B aston.ac.uk...

 John Fletcher wrote:


 Has anyone done the Boost configuration for the DM compiler?

 Thanks

 John


 I am setting about doing this from scratch, and have the structure








set

 up, but has anyone ever run configure test for Boost and could tell








me

 how to do it?  I cannot find anything on the Boost sites about how








to do

 this.

 John Fletcher






 I note the assessments above.  I spent some time on this yesterday and can
 report the following results.

 First of all, how to configure boost.  The advice in the boost community


is

 basically

 1.  run ./configure on Unix.
 2.  read the documents.  These do not fully correspond with the actual


test

 programs.

 so I am back to trial and error.

 There is a file called boost\config\select_compiler_config.hpp and I have


added

 to it the following section

 #elif defined __DMC__
 //  Digital Mars put in BEFORE MPW as DM also defines __SC__ (Symantec)
 //  for backward compatibility but I have NOT tested that.


 which sets up dmc.hpp for Digital Mars.  Note that there is another


compiler

 called MPW which also defines __SC__ so it is important to get in before


that

 one.

 After that the question is what to put into dmc.hpp.

 There are two programs to help with this.

 boost/libs/config/test/config_info.cpp
 boost/libs/config/test/config_test.cpp

 Status:

 config_info.cpp will not compile with DM8.30 owing to lack of <iostream>
 It will compile and run with DM8.30 +STLport and provides some useful
 configuration information.

 config_test.cpp will not compile with DM8.30 + STLport
 or with DM8.31.1 + STLport and the compilation error does not change.

 I am currently investigating this to pin it down.  It could be solveable


by a

 BOOST config choice (although I think I have tried most of those already)


or it

 could need a change to the DM compiler.

 If anyone else is interested I will share the files mentioned above.

 John Fletcher
Oct 17 2002
prev sibling parent reply "Walter" <walter digitalmars.com> writes:
"John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
news:3DAE7694.C9958229 aston.ac.uk...

 config_test.cpp will not compile with DM8.30 + STLport
 or with DM8.31.1 + STLport and the compilation error does not change.

 I am currently investigating this to pin it down.  It could be solveable


by a

 BOOST config choice (although I think I have tried most of those already)


or it

 could need a change to the DM compiler.

 If anyone else is interested I will share the files mentioned above.


I'm always interested in any concise, succint bug reports!
Oct 17 2002
parent reply John Fletcher <J.P.Fletcher aston.ac.uk> writes:
Walter wrote:


 "John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
 news:3DAE7694.C9958229 aston.ac.uk...

 config_test.cpp will not compile with DM8.30 + STLport
 or with DM8.31.1 + STLport and the compilation error does not change.

 I am currently investigating this to pin it down.  It could be solveable


 by a

 BOOST config choice (although I think I have tried most of those already)


 or it

 could need a change to the DM compiler.

 If anyone else is interested I will share the files mentioned above.


 I'm always interested in any concise, succint bug reports!


Walter

I was trying to get to that, but when I extracted the code where the compiler
error message is, it went away, so I think it must be something to do with
where the template was being called from, which was not given.  Is there any
way to get that information?

gcc does it with "instantiated from ...." and this provides a chain back.

I think in this case the call is coming from something to do with output, so
that may take me into <iostream>.  At the moment this is preventing me from
compiling the configuration testing program which comes with Boost!

John
Oct 18 2002
parent "Walter" <walter digitalmars.com> writes:
"John Fletcher" <J.P.Fletcher aston.ac.uk> wrote in message
news:3DAFDA1B.53741675 aston.ac.uk...

 I was trying to get to that, but when I extracted the code where the


compiler

 error message is, it went away, so I think it must be something to do with
 where the template was being called from, which was not given.  Is there


any

 way to get that information?

 gcc does it with "instantiated from ...." and this provides a chain back.

 I think in this case the call is coming from something to do with output,


so

 that may take me into <iostream>.  At the moment this is preventing me


from

 compiling the configuration testing program which comes with Boost!


The only way I've found to do it is to copy all the files to a separate
directory, and then whack away at them until the smallest file is there that
reproduces the problem. I frequently will merge all the include files in by
compiling with -e -l.
Oct 18 2002
prev sibling parent John Fletcher <J.P.Fletcher aston.ac.uk> writes:
Matthew Wilson wrote:


 I would suggest that would be the case also. The last time I looked, they
 don't have a facility for disabling the boost namespace (which stlsoft does
 have, ;)), so any out-of-class template-class member functions are likely to
 give nasties


I think this accounts for the problems I am having.  I have sent my work to this
point to Matthew as he asks (see below) and will work on other things for the
time being.

John
Oct 17 2002