On Thu, 28 Apr 2022, Richard Biener via Gcc-patches wrote:
> We are eventually ICEing in decimal_to_decnumber on non-decimal
> REAL_VALUE_TYPE that creep in from uses of build_real (..., dconst*)
> for DFP types. The following extends the decimal_to_decnumber
> special-casing of dconst* to build_
We are eventually ICEing in decimal_to_decnumber on non-decimal
REAL_VALUE_TYPE that creep in from uses of build_real (..., dconst*)
for DFP types. The following extends the decimal_to_decnumber
special-casing of dconst* to build_real, avoiding the bogus REAL_CSTs
from creeping into the IL and mod
On Wed, 27 Apr 2022, Richard Biener via Gcc-patches wrote:
> Is this the correct strathegy to deal with this problem? Would
> it be valid to just set ->is_decimal in build_real based on
Just setting ->decimal isn't correct; that signifies that ->sig stores the
number in decimal128 format (host-
We are eventually ICEing in decimal_to_decnumber on non-decimal
REAL_VALUE_TYPE that creep in from uses of build_real (..., dconst*)
for DFP types. The following fixes the single occurance that matters
for the testcase in the PR by instead using build_real_truncate.
Bootstrapped and tested on x86
