On 6/28/25 3:17 PM, Florian Weimer wrote:
* Jason Merrill:
Since r10-6069[1] we control the call to __cxa_finalize with
DEFAULT_USE_CXA_ATEXIT, so there's no need to also declare it as a weak
reference; if the target has __cxa_atexit, it must also have __cxa_finalize.
I expect that most targets do not need __cxa_finalize. They can run the
registered destructors directly, without waiting for a call to
__cxa_finalize.
AFAICT glibc still uses __cxa_finalize to run destructors on dlclose;
at least that's what its comment says.
Alternatively, we could replace -lc with
--push-state --no-as-needed -lc --pop-state
which is less likely to break obscure uses, but could result in libc
dependencies that are actually unneeded. And some of the obscure uses have
been broken since GCC 10.
I think with your patch, libc will is always linked in as well, assuming
that it's the source of the __cxa_finalize definition.
True, so that's not a difference for __cxa_atexit targets.
The difference between the --push-state solution and the patch is that
on glibc-like targets, a conditional branch during startup is avoided.
And it will be necessary to link with --nostartfiles in more cases (or
provide a dummy definition of __cxa_finalize).
Or properly configure the compiler with --disable-__cxa_atexit.
The point of my patch is that this should be (and already halfway is) a
configure-time choice, not something that can change from one TU to another.
Jason
