FWIW I would vote for a) "document gpgrt-config in detail" and suggest using pkg-config (variant) for direct invokations. There seems to be little benefit in investing effort/complicating gpgrt-config when pkg-config works fine.

Bootstrapping is an issue. Recall that pkg-config is not a simple program but requires the use of glib (which depends on libffi, libmount, libpcre) - catch-22. Makes building GnuPG on AIX not actually easy.

In T6288#165435, @werner wrote:

Bootstrapping is an issue. Recall that pkg-config is not a simple program but requires the use of glib (which depends on libffi, libmount, libpcre) - catch-22. Makes building GnuPG on AIX not actually easy.

Hello,

that seems to be unrelated to this bug report. Afaiui gpg-* bootstrapping is solved by AM_PATH_GPGME et al. (Which use gpgrt-config and do the heavy lifting of find the correect --libdir option for gpgrt-config .)

pkg-config most popular implementation nowadays is pkgconf which does not have the expensive glib dep and should be easy to bootstrap.

cu Andreas

Something like this?

@ametzler1 Thanks a lot for your help.

I'm going to add the documentation.

I consider again, and I realize that the options are not exclusive; We could also support a simple invocation with no --libdir or PKG_CONFIG_* variable, and then, use it in autoconf macro to determine, for next invocation of gpgrt-error with --libdir. That is, factor out the code from gpg-error.m4, copy it to gpgrt-config as a fallback. In future, gpg-error.m4 can remove the code (after we make sure new gpgrt-config availability).