Front page | perl.perl5.porters |
Postings from November 2016
On Sat, Nov 19, 2016 at 7:51 AM, Niko Tyni <ntyni@debian.org> wrote: > On Fri, Nov 18, 2016 at 02:24:23PM -0800, Pip Cet via RT wrote: >> That patch fixes things for ldinf, but doesn't appear to do anything for ldnan. > I only patched the obvious bugs. I don't see randomness (or even non-zero > high bytes) in ldnan here so I didn't touch that part. I think that's just luck, though. The code generated by gcc here just leaves the high six bytes alone, so it's whatever was on the stack when main() was called—which doesn't just depend on the compiler, it depends on the precise build of libc and any preload libraries. glibc, at least, does quite a few things before finally calling main(). > But yes, ldnan should probably get the same treatment. Patch attached. I think that's the best we can do, given the rather unruly way some compilers appear to behave.Thread Previous | Thread Next