Getting the names from libc sounds ideal, but i don't think there's any way to do that with musl, which is what i'm trying to cross compile to. That's why i attempted a more universal solution in my patch, but perhaps there would be interest in musl to add sigabbrev_np instead. As for a fallback, shouldn't the goal be something more similar to my attempt rather than the existing code? Silently building a broken version of dash when cross compiling isn't ideal. Thanks Henrik On 2024-04-27 13:47, Herbert Xu wrote: > Henrik Lindström <henrik@xxxxxx> wrote: >> They were previously generated at buildtime by mksignames.c, but that >> approach had two flaws: >> 1. The signal names were generated for the host system rather than the >> target system, resulting in broken cross-compiled builds. >> 2. The SIGRTMIN and SIGRTMAX macros are usually implemented as >> function calls and can only be surely known at runtime. >> >> The new implementation has been tested to generate identical signal names >> as before on these systems: >> * Debian 12 (glibc, odd number of realtime signals) >> * Alpine 3.18 (musl, even number of realtime signals) >> * FreeBSD 14 >> >> Signed-off-by: Henrik Lindström <henrik@xxxxxx> > > Now that glibc has sigabbrev_np we should switch to using that > on Linux, perhaps with the existing code as a fallback. BSD > has always had ways of getting the signal name, though it may > not be very portable so we'd need different flavours if people > cared enough to add them. > > Thanks,