Bug status and Stage 3

[Tobias Burnus]

Hello,

as the Topic of #gcc nicely states: "Trunk - Stage 3 Ends October 31 ||
Still 215 regressions in 4.2, 130 in 4.3 and 179 in 4.4"

Thus we have still two weeks to fix bugs for 4.4 and then GCC will be in
the regression-only stage.


Current regressions:

37821 - [4.4 Reg.] CPP include path problem
  As Daniel F. has currently no time, that's probably something for me

37754 - [4.4 Reg.] READ I/O Performance
  13% slower formatted reads than 4.3. I think Jerry has a look.
  However, for performance regressions, having more eyes can never
  harm as everyone finds different bottlenecks

36463 [4.4 Reg.] ICE on valid code
  This is not a true regression but something unimplemented, which
  now fails latter (before there was a bogus error message). The change
  was due to a bug fix. Still, we should try to get this right.


Fortran95-tagged bugs: 16 (of which 37787 is already fixed for the trunk).
Some need ABI changes, but we could still try to get the number down a bit
http://gcc.gnu.org/bugzilla/showdependencytree.cgi?id=32834&hide_resolved=1

Other bugs: One should do some triaging as we might have missed to tag
some bugs; and maybe we can also fix some non-F95 bugs. See also "Bug
Bashing" at http://gcc.gnu.org/wiki/GFortran


In my opinion, the gfortran shape is quite good as we have hardly any
regressions and the F95 list not too long; the number of new features is
slower than hoped/planned, but that's unfortunately normal.


Tobias


PS: The next GCC 4.4 status report by Mark will probably posted soon -
the latest end of this month; the old status was:
http://gcc.gnu.org/ml/gcc/2008-09/msg00221.html (9 September)


PPS: To look at other compilers: g95 now supports Fortran 2003 Coarrays
(on shared- and distributed-memory systems). Having some stub
implementation (single image) might be also a useful goal for gfortran
4.5. The other big features I would like to see is OOP, possibly with
derived-type kind- and length parameters, and IEEE. Let's see how far we
can get in 4.5. Maybe with 4.6 gfortran will be - modulo bugs - a full
F2003 compiler?


PPPS: "The Fortran 2003 Handbook" by Adams, Brainerd, Hendrickson,
Maine, Martin and Smith (Springer, 2008) is supposed to be available
"October 16, 2008" in North America (and in the UK?) and in November in
the rest of the world. Amazon.com already offers "look inside" as does
books.google.com and it is available in PDF form at
http://www.springerlink.com/content/m25273/ (if you have a subscription;
I don't but the table of contents and the preface are freely available).
The book very much looks like the Fortran 2003 standard (same structure
of the chapters) and is supposed to be a more readable version of the
standard with more examples and given reasons for the choices. (See
preface.)