I wrote:
> So I pushed that, and not very surprisingly, it's run into some
> portability problems. gombessa (recent OpenBSD) reports
> ! nm -A -g -u libpq.so.5.15 2>/dev/null | grep -v '_eprintf\\.o:' | grep -e abort -e exit
> libpq.so.5.15:__cxa_atexit
After a few more hours, all of our OpenBSD animals have reported
that, on several different OpenBSD releases and with both gcc
and clang compilers. So at least it's a longstanding platform
behavior.
More troublingly, fossa reports this:
! nm -A -g -u libpq.so.5.15 2>/dev/null | grep -v '_eprintf\\.o:' | grep -e abort -e exit
libpq.so.5.15: U abort@@GLIBC_2.2.5
Where is that coming from? hippopotamus and jay, which seem to
be different compilers on the same physical machine, aren't showing
it. That'd lead to the conclusion that icc is injecting abort()
calls of its own accord, which seems quite nasty. Lacking an icc
license, I can't poke into that more directly here.
Perhaps we could wrap the test for abort() in something like
'if "$CC" != icc then ...', but ugh.
regards, tom lane