Need a global GCC maintainer ruling on atexit

Richard Henderson rth@redhat.com
Fri Dec 29 01:18:00 GMT 2000


On Thu, Dec 28, 2000 at 08:07:00PM -0800, Michael Sokolov wrote:
> 1. Provide atexit in libiberty (or elsewhere) for systems that don't have it.
> 2. Change programs needing atexit functionality to use xexit/xatexit.
> 3. Conditionally #define exit to xexit and atexit to xatexit on systems that
>    lack atexit.

Since libiberty provides xatexit (for reasons explained elsewhere),
clearly Nr 1 is not correct.

> Right now the atexit problem occurs only in fixinc, and the fixinc
> maintainer Bruce Korb said that he prefers the 3rd solution.  However,
> I think it is just a happenstance that the atexit problem currently
> occurs in fixinc and not somewhere else ...

Ideally I think we'd have system.h do Nr 3 when needed, but at present
that would cause problems with canadian crosses since we do not link
genfoo against libiberty.  So, failing that, I don't see any harm in
restricting this sort of fix to fixinc.


r~


More information about the Gcc-patches mailing list