This is the mail archive of the
mailing list for the GNU Fortran project.
GNU Fortran IRC meeting planning: When? Agenda?
- From: Tobias Burnus <burnus at net-b dot de>
- To: gfortran <fortran at gcc dot gnu dot org>
- Date: Thu, 23 Sep 2010 19:24:24 +0200
- Subject: GNU Fortran IRC meeting planning: When? Agenda?
as suggested about 10 days ago by Jerry and seconded by Paul, we should
do another planning/design review meeting in gfortran's IRC channel
(irc.oftc.net, channel #gfortran).
To decide about when and where, I have setup a poll at
Please vote for your favourite time and date.
For the agenda, I think we have three blocks:
a) Larger items for Stage 1
b) Tasks for 4.6.0
c) Tasks for 4.7.0
Some thoughts for the agenda of the meeting - feel free to suggest more
(or comment, though the real discussion will be on IRC)
(a) As Stage 1 ends on October 27, we still have a bit more than a month
for medium-size projects. I think implementable and thus worth to
discuss (or get status updates) are
- REAL(16) support for __float128 systems
- Deferred (allocatable) character lengths (at least parsing) [PR45170]
- Additions to ASSOCIATE
- OOP fixes?
- Larger fixes such as allocatable components of allocatable scalars
- Whole-file fixes (PR 45756, PR 44945, and some more)
- Some smaller new features (e.g. from F2008)?
- CPP: Add some target support, e.g. Kai's patch for PR 42954
(b) Tasks until the release (can be done in Stage 1 - or later)
- There are 8 regressions, two seem to be target specific, one is a
middle end one, three seem to be mostly fixed - thus, there are at least
two untackled ones left.
- The number of smaller issues (wrong code ,ice on valid/invalid code)
has be gone high as most work concentrated on Stage 1 items - we should
try to get it down
- Testing (correctness, performance): Running checks oneself, asking
(c) Larger tasks:
- Array descriptor update: Something to get ready for 4.7; needs some
work - so far only Paul was brave and worked on it. Besides the basic
support, we need to think about which fields should be included -
includes thinking about CLASS(*) and about DT kind/len parameters
- Memory tracking:* Diagnostics - but also correctness (let DEALLOCATE
fail for nonallocated memory like named TARGETs)
- FINAL:* Daniel wants to work on this ...
- OOP - arrays and unlimited polymorphism
- Coarrays - going truly multi-image
- IEEE - most hated item, least wanted item - but feature most other
- Derived-type type parameters for kind (Sa's patch*) and for len(gth)?
- Submodules & speedup of module reading [revamp of the format?]
- libcpp - going away from -traditional-cpp and getting more
TARGET_*_CPP_BUILTINs to work (cf. PR 42954 above)
- Reallocate on assignment (much missed item)
* * *
- Fortran 2008 is now an ISO standard (accepted (stage 60.00) but not
- TR 29113 (Further Interoperability of Fortran with C) should have a
draft in October and should return from a PDTR ballot in April - thus at
some point next year the question about implementing it will pop up. At
least for "dimension(..)" and "type(*)" there seems to be some demand
from MPI vendors (and from the MPI Forum for MPI 3) Cf.
- Looking at the F2003 and F2008 implementation status, the major
unimplemented items seem to be Fortran 2003 - though coarrays and
submodules will also require some work. Cf.
- For old plans and action items, see http://gcc.gnu.org/wiki/GFortran46
* Cf. http://gcc.gnu.org/wiki/GFortranPatchTracker