GCC 3.0 Status - Fortran.

[Toon Moene]

Lectoribus Salutem,

The following is a list of the current Fortran and Fortran-related bug
reports in GNATS:

Fortran/

542  GNU Fortran doesn't support direct access files bigger than 2Gbyte.

     This is a change-request, no regression; needs some serious hacking 
     in libf2c with accompanying autoconfigury.

947  Data statement initialization of an array element whose index is a
     INTEGER*2 constant hits an assert in the Fortran Frontend.

     A long time shortcoming of the Fortran Frontend - no regression.

1832 List directed I/O hangs on string bounds overrun.  -fbounds-check
     erroneously doesn't detect this.

     Has been wrong always - no regression.
     Requires some serious libf2c hacking.

2031 Internal Compiler Error in g77.f-torture/compile/20000511-2.f on 
     mips-sgi-irix6.5.  This was fixed on the GCC 3.0 branch with this 
     patch:

     http://gcc.gnu.org/ml/gcc-patches/2001-04/msg01092.html

     The fix was deemed not suitable for the trunk.

2153 Statement function argument namespac overlaps subroutine namespace.

     The compiler is correct.  However, the error message is not
correct, 
     hence listed as a change-request - not a regression.

2244 Source with an incredible number of ASSIGNED GOTO's manages to
     draw the following Internal Compiler Error when compiled with -O2
-g
     on i686-pc-linux-gnu.

     2244.f:1323: Internal compiler error in make_edges, at flow.c:1152

     This is a regression from GCC-2.95.3, so the GNATS entry is marked
     HIGH priority.

2661 g77.f-torture/execute/970625-2.f fails on mips-sgi-irix6.5 with 
     -mabi=64 -O2

     Fails since the change-over to BYTE offsets in SUBREGs on the
trunk.
     [ So not relevant for 3.0 status ]

2667 Source draws error:
     2667.f:3557: Internal compiler error in
reload_cse_simplify_operands, 
     at reload1.c:8352
     when compiled with  -O -msse -c.

     This is not a regression, because -msse was not known in 2.95.x.

libf2c/

941  Configuration of libf2c fails when cross-compiling.
     Improper use of auto-host.h file.  No regression.

optimization/

2463 A Fortran source failing on optimization -Ox with x >= 2.

     Not a regression.  This fails on all compilers I have on my Debian
     2.2 Alpha system: 2.95.2, 2.95.3, 3.0 and 3.1 as-of-last-week.
     Installed as g77.f-torture/execute/20010430.f on the trunk.

<HINT MODE=SUSTAINABLE-BUSINESS-MODEL>
The items marked "change-request" will be tackled when I have managed to
secure support contracts from one of the larger Fortran using
conglomerates in the world.  Recently, I learnt that Boeing installed an
MSC cluster running NASTRAN for doing large scale Computational Fluid
Dynamics calculations.  Another obvious candidate is the Centre Europeen
de Recherche Nucleaire, whose facilities are so large that they're
spread over two countries.

Or I have to acquire some serious amount of spare time.

Whichever comes first.
</HINT>

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)