gcc 2.8.1

Olaf Titz (olaf@bigred.inka.de)
Mon, 20 Jul 1998 02:43:31 +0200

...is weird. It produces a rock solid Linux 2.1.108 kernel (9.3MB
source, bzipped) but miserably fails on poor little (874k) INN 2.0.
More precisely, the 980719 snapshot I compiled with gcc 2.8.1 on
aforementioned Linux (K6), to be run on a Linux 2.0.35 (i486) system,
both Debian 1.3 (libc 5) based, reproducably generates broken overview
stuff (with makehistory and expireindex, from scratch). This results
in XOVER getting (semi-?) correct results, followed by ARTICLE giving
bogus ones, while strace shows that the ARTICLE command tries to
access the wrong (timehash) file.

INN compiled with gcc 2.7.2.3 is OK. Any insight? I've CCd this to the
Linux and INN lists because perhaps
- INN people know that INN does some exotic bit-munging which breaks
gcc, or
- Linux has subtle inconsistencies between 2.0 and 2.1 which makes
stuff compiled on 2.1 fail on 2.0?

Both would be serious bugs, either in gcc or Linux (or even INN?). I
suspect gcc but does INN really contain code which is so critical, or
is gcc just still broken and unfit for public consumption like 2.8.0 was?

olaf

PS. I compiled Mozilla (16MB, gzipped) with 2.8.1 and it works (modulo
lesstif bugs). Nobody claimed that _that_ is clean code.

PPS. gcc 2.8.1 _seems_ to run faster than 2.7.2.3, but maybe that's
just me...

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html