On Wed, Dec 08, 2021 at 10:55:41PM +0900, Takashi Sakamoto wrote: > I'm not positively for the patch since it can take developers puzzled due to > the symbol dependency newly introduced in unreleased version of alsa-lib. Shouldn't the version check and local definition avoid that issue - if the version of alsa-lib doesn't have snd_config_load_string() then we'll use a locally defined version of snd_config_load_string() and not depend on the alsa-lib symbol? > It's better to check the version of alsa-lib in Makefile to avoid developers’dole > if we have enough respect to embedded developers, especially forced to work > with legacy userspace. (and it often occurs). Or just avoid using fancy new library features - if we need to limit the test to only building with bleeding edge versions that gets restrictive. TBH this is probably even more painful for people working with enterprise distros than embedded systems, if you're building everything for your target it's not usually too bad to drop in an updated version of something like alsa-lib but if you're using disro binaries it's less idiomatic. Either way it's a barrier though.
Attachment:
signature.asc
Description: PGP signature
