This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
namelist patch [was egcs-1.0 alphaev56-unknown-linux-gnulibc1 test results]
- To: egcs at cygnus dot com, toon at moene dot indiv dot nluug dot nl
- Subject: namelist patch [was egcs-1.0 alphaev56-unknown-linux-gnulibc1 test results]
- From: Richard Henderson <rth at cygnus dot com>
- Date: Tue, 9 Dec 1997 16:25:13 -0800
- Cc: "Harvey J. Stein" <hjstein at bfr dot co dot il>, fortran at gnu dot org
- References: <199712091711.TAA10227@blinky.bfr.co.il> <9712092233.AA02622@moene.indiv.nluug.nl>
- Reply-To: egcs at cygnus dot com
On Tue, Dec 09, 1997 at 11:33:06PM +0100, Toon Moene wrote:
> > ... so as to avoid trying to build the fortran
> > compiler (because it fails).
>
> Fascinating - it works for me (system type
> alphaev5-unknown-linux-gnulibc1 :-)
It's going to depend on how it is built exactly, and with what version
of binutils. It is quite probable that f771 overflows the .got segment
when using binutils 2.7.0.2 -- later versions can handle multiple .got
segments. If you care, you upgrade. Nothing for it.
> It dies, however, on the following (Fortran code):
>
> namelist/junk/a,b,c
>
> a=1.234
> b=5.678
> c=-123.456
>
> write(10,junk)
>
> stop
> end
Yep. Here's a patch.
r~
Tue Dec 9 16:20:57 1997 Richard Henderson <rth@cygnus.com>
* com.c (ffecom_type_vardesc_): Vardesc.dims is a `ftnlen*'.
Index: com.c
===================================================================
RCS file: /cvs/cvsfiles/egcs/gcc/f/com.c,v
retrieving revision 1.8
diff -u -p -d -r1.8 com.c
--- com.c 1997/11/03 06:07:00 1.8
+++ com.c 1997/12/10 00:16:49
@@ -9921,7 +9921,7 @@ ffecom_type_vardesc_ ()
addrfield = ffecom_decl_field (type, namefield, "addr",
string_type_node);
dimsfield = ffecom_decl_field (type, addrfield, "dims",
- ffecom_f2c_ftnlen_type_node);
+ ffecom_f2c_ptr_to_ftnlen_type_node);
typefield = ffecom_decl_field (type, dimsfield, "type",
integer_type_node);