This is the mail archive of the
libc-alpha@sourceware.cygnus.com
mailing list for the glibc project.
Re: [Craig Metz <cmetz@inner.net>] libc/1184: Latest glibc +toolchain can't really build 386 lib
- To: Horst von Brand <vonbrand@sleipnir.valparaiso.cl>,Andreas Jaeger <aj@arthur.rhein-neckar.de>
- Subject: Re: [Craig Metz <cmetz@inner.net>] libc/1184: Latest glibc +toolchain can't really build 386 lib
- From: Joel Klecker <jk@espy.org>
- Date: Thu, 1 Jul 1999 18:57:53 -0700
- Cc: libc-alpha Mailinglist <libc-alpha@sourceware.cygnus.com>,Craig Metz <cmetz@inner.net>
- References: <199907012335.TAA10003@sleipnir.valparaiso.cl>
At 19:35 -0400 1999-07-01, Horst von Brand wrote:
>Andreas Jaeger <aj@arthur.rhein-neckar.de> said:
>> We've received the append bug report. Any ideas what's broken here?
>
>[CFLAGS -m386 on an i686 gives programs that SIGILL on i386]
>
>The problem is that -m386 schedules for a i386, but generates native code
>(i686 in the case discussed here); you need -march=i386 (generate code for
>i386, and schedule for it too). BTW, -m386 is deprecated.
I have tested it, it seems that only -mcpu=i386 will actually have
the desired effect. With egcs 1.1.2 and gcc 2.95 as of 19990629
configured for i486-linux, -march=i386 or -m386 still end up defining
'i486', '__i486', and '__i486__'.
I don't actually have a library built this way, however, and would
have no way to test it if I had.
--
Joel Klecker (aka Espy) Debian GNU/Linux Developer
<URL:mailto:jk@espy.org> <URL:mailto:espy@debian.org>
<URL:http://web.espy.org/> <URL:http://www.debian.org/>