future merge

[Peter Mazinger]

> 
> On Jul 9, 2011, at 6:08 AM, "Peter Mazinger" <ps.m at gmx.net> wrote:
> 
> > Khem, the master below is unbuildable, fails on epoll.c with your
> config. This epoll.c is the one from master, not mine ;-(

committed a fix to the MIPS problem to future branch, affects though all archs using NPTL or LT new. The problem comes from the headers that provide
different prototypes for the same functions, some are hidden, others visible
(and weak).

There should be some common rules we could agree about, like:
- make attribute_hidden (and lib*_hidden_*) noop for *.o and *.os files
- include only *.oS in the shared libraries (compiled with -DSHARED), this
way hiding functions becomes active only in the shared libs (avoiding jump
relocations), the only bad effect is that it makes build time longer
- does it make sense to have weak + hidden functions?
- define when weak functions should be used (ex: if the same function is
in libc and libpthread, one of them has to be weak to avoid linking failure
- remove duplicate functions from libc and libpthread if the one in 
libpthread just calls the one from libc

other ideas?

Peter
-- 
Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir
belohnen Sie mit bis zu 50,- Euro! https://freundschaftswerbung.gmx.de