Re: could gfortran be tested on Darwin regress builds of 4.1?

Re: could gfortran be tested on Darwin regress builds of 4.1?

[Bill Northcott]

On 30/07/2005, at 1:06 AM, Jack Howarth wrote:
>    Well the gcc4.info for the package containing gfortran in fink  
> has...
>
> Depends: gmp-shlibs (>= 4.1.3-11), cctools (>= 576-1) | odcctools  
> (>= 576-200503
> 27), %N-shlibs, libiconv
> BuildDepends: gmp (>= 4.1.3-11), libiconv-dev
>
> which is a tad confusing since I guess it implies that cctools  
> isn't needed
> to build gcc4/gfortran but is needed to use it. That seems weird  
> since to
> build it and do a make check is to effectively use it.
>

I built a working (builds functioning R and Octave) gfortran with  
Apple's gcc 4 on MacOS X 10.4.1 (Darwin 8.1).

The only thing I needed to add was gmp, which built from the source  
without problems.

The cc-tools in Darwin 8 is 590.  So I think it would be a bad idea  
to install 576!

If I get time, I will try the current code.

Bill Northcott

Re: could gfortran be tested on Darwin regress builds of 4.1?

[Peter O'Gorman]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Bill Northcott wrote:
| On 30/07/2005, at 1:06 AM, Jack Howarth wrote:
|
|>    Well the gcc4.info for the package containing gfortran in fink  has...
|
| The only thing I needed to add was gmp, which built from the source
| without problems.
|
| The cc-tools in Darwin 8 is 590.  So I think it would be a bad idea  to
| install 576!

Um, xcode-2.1 features cctools-590, darwin8 and Mac OS X 10.4.0, shipped
with cctools-576.

I too think that it would be great if Geoff could install gmp on the gcc
tester, but can understand his reasons for caution.

Peter

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (Darwin)

iQCVAwUBQuquO7iDAg3OZTLPAQIIYwP9Hd2I4B4RFNZTx0KdCWfxYz++WE+d3mn3
LYXN1tV7KqwrMKIkm/Nk+DIFg+71FR4/DphJl8MN6IDrHGnPzYuWnlVbX1QAj6UZ
x/wbPp/q+2/iezfSuTmtRvKmzRnN8sFgb0S1Yit3nVOwt+NsNNh1iD+TWN4QO86M
2cwTzjBPff8=
=rd/3
-----END PGP SIGNATURE-----

Re: could gfortran be tested on Darwin regress builds of 4.1?

[Jack Howarth]

Geoffrey,
   Well the gcc4.info for the package containing gfortran in fink has...

Depends: gmp-shlibs (>= 4.1.3-11), cctools (>= 576-1) | odcctools (>= 576-200503
27), %N-shlibs, libiconv
BuildDepends: gmp (>= 4.1.3-11), libiconv-dev

which is a tad confusing since I guess it implies that cctools isn't needed
to build gcc4/gfortran but is needed to use it. That seems weird since to
build it and do a make check is to effectively use it.
              Jack

could gfortran be tested on Darwin regress builds of 4.1?

[Jack Howarth]

     Since gfortran is making such good progress at this point, it
would seem like a really good idea for Apple to add the gfortran build
to its builds on regress. It would make is easier for Mac users to
tell what the expected status is of gfortran on Darwin. Thanks in
advance for considering this.
          Jack

Re: could gfortran be tested on Darwin regress builds of 4.1?

[Geoffrey Keating]

howarth@bromo.msbb.uc.edu (Jack Howarth) writes:

>      Since gfortran is making such good progress at this point, it
> would seem like a really good idea for Apple to add the gfortran build
> to its builds on regress. It would make is easier for Mac users to
> tell what the expected status is of gfortran on Darwin. Thanks in
> advance for considering this.

The regression tester doesn't make such policy decisions; it simply
does '.../configure && make bootstrap'.  Thus, as you see, the
expected status of gfortran on Darwin is that it doesn't get built.

I believe this has something to do with Darwin not having some
libraries that gfortran requires, and possibly also gfortran not being
enabled by default.  I'm hesitant to install those libraries specially
for the regression tester; the tester is not supposed to be testing
optional configurations, it is testing the default build.  It would be
better if gfortran could be changed so that it did not require those
libraries, or if those libraries were provided as part of the tree.