Replying to two emails at once here:
On 30 May 2006 at 10:37, Camm Maguire wrote:
| Greetings!
|
| Dirk Eddelbuettel <[EMAIL PROTECTED]> writes:
|
| > On 26 May 2006 at 22:45, Frederik Schueler wrote:
| > | Hi,
| > |
| > | On Fri, May 26, 2006 at 03:18:35PM -0500, Dirk Eddelbuettel wrote:
| > | > All fine, but who depends on g77? For R, I make'em build-depend on
| > | > r-base-dev, and that now has
| > |
| > | refblas3-dev depends on it, which is a dependency of r-base-dev. Looking
| > | closer, r-base-dev depends on
| > |
| > | refblas3-dev | atlas3-base-dev | libblas-3.so
| > |
| > | and the buildds use to pick the first one by default. But atlas3-base-dev
| > | depends on g77 too...
| >
| > Good work, thanks! Had forgotten about that one.
| >
| > Camm: Would it make sense and/or be possible to 'roll' atlas and blas
| > forward to gfortran? I recently did that with R and it seems to work fine.
| >
|
| Sure, but
|
| 1) is this 'recommended' -- i.e. is one of these options
| temporary or less 'official' than the other?
| 2) what differences do you observe that are required? I.e. is
| this simply a binary abi issue?
I am not sure what changed between g77 and gfortran, but I believe g77 will
go the way of the dinosaur as gcc 4.* only produces only gfortran.
On 30 May 2006 at 10:39, Camm Maguire wrote:
| Greetings!
|
| Dirk Eddelbuettel <[EMAIL PROTECTED]> writes:
|
| > On 26 May 2006 at 22:45, Frederik Schueler wrote:
| > | Hi,
| > |
| > | On Fri, May 26, 2006 at 03:18:35PM -0500, Dirk Eddelbuettel wrote:
| > | > All fine, but who depends on g77? For R, I make'em build-depend on
| > | > r-base-dev, and that now has
| > |
| > | refblas3-dev depends on it, which is a dependency of r-base-dev. Looking
| > | closer, r-base-dev depends on
| > |
| > | refblas3-dev | atlas3-base-dev | libblas-3.so
| > |
| > | and the buildds use to pick the first one by default. But atlas3-base-dev
| > | depends on g77 too...
| >
| > I realized that we were looking in the wrong corner. (Indirect)
Build-Depend:
| > on gcc et 3.4 are irrelevant as the default gcc et al are alread 4.0.* in
| > testing and unstable.
| >
| > So the bug really is that gfortran is stricter on the code and fails.
| >
|
| blas/lapack/atlas code, or R code?
|
An R add-on package from the CRAN archive fails under gfortran.
AFAICT blas/atlas/lapack should Build-Depends: and Depends: on gcc which
these days is version 4.0. Unless, of course, you have reasons to depend on
the older 3.4 release. But even if you have reasons, maybe these shouldn't
get exported to users of these packages now that Debian made 4.* the default
version of gcc et al.
Thoughts?
Dirk
--
Hell, there are no rules here - we're trying to accomplish something.
-- Thomas A. Edison
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
