This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Should this be an error (SPEC2000 source)?
- To: gcc at gcc dot gnu dot org
- Subject: Should this be an error (SPEC2000 source)?
- From: Brad Lucier <lucier at math dot purdue dot edu>
- Date: Thu, 27 Jan 2000 12:33:03 -0500 (EST)
- Cc: lucier at math dot purdue dot edu
In the source code for the 176.gcc test of SPEC2000 there is the line in
bc-typecd.def:
DEFTYPECODE (SFcode, "SF", SFmode, SFtype)
that leads to the compilation error with the 20000121 development
tree (built on alphaev6-unknown-linux-gnu), compiled with -O3
-ffast-math -mcpu=ev6:
In file included from bc-emit.c:948:
bc-typecd.def: In function `bc_emit_instruction':
bc-typecd.def:17: `SFtype' is promoted to `double' when passed through `...'
bc-typecd.def:17: (so you should pass `double' not `SFtype' to `va_arg')
And, when the SFtype is changed to DFtype (which I believe to be double),
everything is built fine, the output passes, etc.
My question: Should this problem really be an error that stops
compilation? Or should it be a warning that allows compilation to
proceed and will (hopefully) generate correct code.
Brad Lucier