IRC Meeting and #gfortran

[Steve Kargl]

On Sun, Nov 05, 2006 at 11:05:02AM -0800, Jerry DeLisle wrote:
> 
> I think we had a fairly successful meeting on IRC yesterday.  
> Unfortunately, I had my session logging turned off.
> 
> However, we did create a #gfortran channel on oftc.net.  So we can plan 
> future meetings similarly.
> 
> Thanks to all who participated.
> 

I too regret not logging the IRC meeting.  However, for the record
Jerry, FX, Tobias, Janne, Brooks, and myself participated in the
discussion (did I miss someone?).  I'd particularly like to thanks
FX, Tobias, and Janne for staying up well past a normal bedtime!
The meeting covered (gang, if I mis-speak, please correct me):

1) Symbol version.  We want to get Janne's (and FX's?) patch
   tested and committed to trunk.  I'm bootstrapping it now on
   i386-*-freebsd.  This patch introduces a -fall-leading-underscore
   option (temporarily?).  It will be possible to remove this
   option once ISO C Binding is functional.

   We also discussed and decided the symbol versioning should be
   back ported to 4.2.  This is probably considered to be a new
   feature, so we'll need to have Mark Mitchell's approval.  I'll
   note that I (we?) feel the benefits of symbol versioning outweigh
   any warts (ie., the new option).

2) Documentation.  I mentioned that I thought we should have a "Processor
   Dependent Behavior" chapter.  The description of the environmental
   variables would become a subsection.  The intrinsic modules would
   be documented here as well as things like default unit numbers for
   stdin and stdout.

   Brooks (whose done some recent work on the docs) mentioned that 
   he felt the manual could use a re-organization.  I agreed with 
   him and asked if he could post an outline of what he had in mind.

   In the end, I think it's safe to say that we all agreed the docs
   could use some work.

3) Intrinsic modules.  We probably needed pault's input and questions
   for this discussion.  FX answered a few questions I had, but truthly
   I can remember neither the questions nor answers. :)  It appears 
   he has plan and I suspect he'll discuss various aspects of the plan
   on the fortran@ list.

4) ISO C Binding.  We know that Chris has a patch caught up in the
   FSF assignment process.  (Have I ever mention how much I dislike
   the assignment process?)  FX mentioned that the patch is large
   and touches a lot of files.  He then broached the topic of creating
   a svn branch to expedite the integration of the patch.  The idea
   is that the ISO C Binding patch would be committed to the branch
   without a formal review.  We then as a group fix any unforeseen
   gotchas.  This would save Chris the headache of receiving several
   review comments (some would clearly be redundant because I'm 
   fairly certain that we'll all spot the obvious issues) and resubmitting
   an enormous patch to the list.    

   One side effect of creating a gfortran-sandbox branch is that
   complicated F2003 projects can be conducted on the branch (ie.,
   submodels and some of the OO features.  In fact, ISO C Binding
   is a feature that probably could have benefitted from development
   on a branch where many eyes could see the code.

5) Record markers.  There was some mention of the record marker 
   issue with gfortran and compatibility with other compilers.
   Thomas Koenig has taken on the bugzilla PR and appears to be 
   making progress.  I think we should make the Intel record and
   subrecord method the default on gfortran with an option to
   permit the current method for backwards compatibility.  Note
   this change will by default break backwards compatibility 
   with previous versions of gfortran; however, I think the 
   benefits of compatibility with many of the commercially 
   available compilers outweighs the merits of backwards compatibility.

Well, that's about all I can remember and it's time to start preparing
dinner.

-- 
Steve