Best practice for linking against libatomic?

Simon Richter <[email protected]> Mon, 12 Jan 2026 19:12:57 +0900

Newsgroups	gmane.comp.gcc.help
Message-ID	<[email protected]>

Hi,

I have a pile of build failures on RISC-V because the program uses 
atomics but doesn't link against libatomic.

A lot of programs seem to notice the need for libatomic when they first 
use some construct that isn't inlined on x86_64 -- some then add a 
configure time test if it is actually needed, for good measure.

The "correct" approach would probably be to test if libatomic is needed 
for all data types or at least sizes that are then used in the actual 
program, but that's a game of whack-a-mole.

The less correct, but lazy approach is to always link against libatomic 
with --as-needed (which assumes the GNU linker).

Is there a recommended way of handling this for package maintainers, or 
should this really be done at system level (e.g. by hacking the specs)?

    Simon

OpenPGP_signature.asc (application/pgp-signature, 488 B)