This is the mail archive of the
mailing list for the glibc project.
ffs vs __ffs the gcc saga continues
- To: libc-alpha at sources dot redhat dot com
- Subject: ffs vs __ffs the gcc saga continues
- From: Raymond Burns <rayburns at eskimo dot com>
- Date: Mon, 12 Mar 2001 11:11:43 -0800 (PST)
Playing with gcc cvs (gcc version 3.1 20010309 (experimental)) against
glibc (cvs Mar 10 12:06 local time) and others. After munging the
configure scripts to allow gcc 3.1 I found the compiler develops a
personal problem with calls in the libc tree to ffs (with a lousey
Replacing calls to ffs with ffsl or __ffs in about 5 places in the
tree caused it to compile and is running on the development box (chose
__ffs). Test code I wrote using ffs, ffsl and ffsll worked for int and
long int , long long int gave odd results.
I'm not sure if this is a symptom of weak aliasing , I'm also not sure
that this is any sort of a fix. It does allow me to exorcise gcc's
contributions to various code compiled with gcc-3.1 using libc-2.2.2
and kernel 2.4.2 (which is another story).
this is an intel 586 box (can only afford the cheap stuff)
Tracking down why the compiler complained led me into a bunch of
#define and #undef's that i didn't trust myself to monkey with
(multi platform source is a Lot of fun)
My gut say's gcc is still very broken and that the glibc tree is
relatively clean, I'm watching for changes to gcc.
Enlightened insight from this list is helpful. details upon