On 11/2/21 2:00 PM, Eric Molitor wrote: > Attention should be paid to the change to portability.h. I believe this is the > correct way to expose memrchr for glibc. I've tested it and it works with > glibc > but I primarily use MUSL.
I'm uncomfortable trying to guess which combinations of const and restrict a nonstandard header prototype is going to have from version to version and library to library. That's why I did timer_create_wrap() with a #define from the old name instead of just implementing something to match the header that would bind first before going out to libraries: I don't trust any header with "const" or "restrict" in it... and the #define from the old name is breaking android NDK because when they DO get around to #including the header it renames the prototype to the new function. And the stanza is triggering on the NDK because they don't have -lrt. So if we don't -lrt on the probe gcc fails, and if we DO clang fails. Bravo. Eh, probe without -lrt because it's really a gcc bug: libgcc_eh.a should not be making thread calls in non-threaded programs. This means we always wrap on glibc, and I'm ok with that. Rob _______________________________________________ Toybox mailing list Toybox@lists.landley.net http://lists.landley.net/listinfo.cgi/toybox-landley.net