[Bug tree-optimization/43716] [4.6/4.7/4.8 Regression] Revision 158105 miscompiles doduc.f90

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43716

--- Comment #42 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2012-12-07 13:40:17 UTC ---
> Re-adjust target milestone.  Please somebody revisit the regression status
> and fill in known-to-work/fail fields.  ISTR it only fails on darwin.

(1) doduc.f90 is correctly compiled on x86_64-apple-darwin10 (r194291) with the
options I have tried (typically -fprotect-parens -Ofast -funroll-loops
-ftree-loop-linear -fomit-frame-pointer -fwhole-program -flto, no valgrind
error with -O3 -ffast-math, no error with -fsanitize=address).

(2) Per comment #30, this PR has been reopened for x86_64-unknown-linux-gnu.

(3) Per comment #40:

> tree if-conversion happily executes both arms of the conditional
> unconditionally with -ffast-math, ...

so compiling with -ffpe-trap=invalid traps for results discarded later.