https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93727
--- Comment #56 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> --- (In reply to dave.anglin from comment #55) > On 2026-07-15 12:09 p.m., jvdelisle at gcc dot gnu.org wrote: > > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93727 > > > > --- Comment #54 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> --- > > (In reply to dave.anglin from comment #52) > >> On 2026-06-16 8:47 p.m., jvdelisle at gcc dot gnu.org wrote: > >>> Well we can XFAIL it OR I rework building the float out of the hex digits > >>> more > >>> manually and avoid using strtof. It will take a while to do the latter. > >> > >> It would be great if we could avoid using strtof but we also need to avoid > >> using > >> strtod and strtold. Is there any chance of using strtoflt128 from > >> libquadmath? > >> The __float128 type is a synonym for "long double" on HP-UX and strtoflt128 > >> appears > >> to handle hexadecimal numbers. > > > > Is there a define I can use to determine when building for HP-UX? > > __hpux__ > > Also, is there a compile farm machine you would recommend I use for testing > > on > > this? > > Not that I'm aware off. We could do our own strtold fairly easily and avoid dependency on whether a platform has C99 compliant hex float interpretation in the srtold completely. Right now, using libquadmath is a user option during configuration so I dont want libquadmath to be a dependency.
