On Tue, Jun 1, 2010 at 12:49 AM, Dr. David Kirkby
<[email protected]> wrote:
> The fortran package in Sage is really weird.
>
> 1) The 'src' directory does not contain unchanged upstream code, but has a
> command called 'sage_fortran' burried well down in the directory structure,
> which is a binary of g95, version 4.0.3
>
> 2) But at the top level there is another command called 'sage_fortran' which
> is a perl script.
>
> 3) Most packages, when they get patched, use the convention of adding .p0,
> .p1, .p2 ...etc to the package number. The 'fortran' package however seems
> to use that convention (SPKG.txt starts at patch level 8), but then switched
> to using the date. The SPKG.txt shows:
>
>
> === fortran-20100428  Harold Gutch, 28th Apri 2010) ===
>  * trac 8715 -- fortran-20100118 ignores SAGE_FORTRAN on Linux
>
> === fortran-20100117 (William Stein, Jan 17, 2010) ===
>  * Removed the two linux g95 binaries, and *require* that gfortran be
> installed
> on Linux.
>
> === fortran-20071120.p8 (William Stein Sept 24 2009) ===
>  * improved 64-bit OS X 10.6 detection
>  * The g95 binaries were downloaded from the very nice site:
>    http://ftp.g95.org/
>  * I changed the name of the executable in the g95-install/bin/
>   directory to sage_fortran in each case.
>
> I can see a point of coding a date if its a CVS snapshot, but it's not
> obvious to me why William changed from fortran-20071120.p8 to
> fortran-20100428.
>
> As far as I can see, there appears to be version 4.0.3 of g95 and version
> 4.3.2 of gfortran buried down somewhere in the directory structure. Do we
> really need both g95 and gfortran? If not, should the older 'g95' be
> removed?
>
> So what name should I used if I update this? Some of the many possible
> options might be:
>
> 1) If 'g95' could be removed, call the package fortran-4.2.3, since I
> believe the binaries are all version 4.2.3. Later we append .p0, .p1 etc as
> updates are made.

g95 can be removed.  It hasn't because nobody has got around to it.

The only binaries that matter are the gfortran ones for OS X.

 -- William

>
> 2) If 'g95' can't be removed, so there must remain a mixture of g95 4.0.3
> and gfortran 4.2.3, we call the fortran package fortran-1.0. Then we append
> .p0, .p1 etc in future.
>
> 3) I use the date, as has been recently done, in contravention of the
> convention in the developers guide. (In which case the developers guide
> should be updated to reflect the 'fortran' package uses a totally different
> convention).
>
> 4) Use a pseudo-random number generator to pick a version for the package!
>
> 5) Some other convention.
>
> To me, (1) or (2) seems most logical.
>
> Thoughts?
>
> Dave
>
>
> --
> To post to this group, send an email to [email protected]
> To unsubscribe from this group, send an email to
> [email protected]
> For more options, visit this group at
> http://groups.google.com/group/sage-devel
> URL: http://www.sagemath.org
>



-- 
William Stein
Professor of Mathematics
University of Washington
http://wstein.org

-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to