linux-mips
[Top] [All Lists]

Re: MIPS symbol versioning patches

To: Ralf Baechle <ralf@oss.sgi.com>
Subject: Re: MIPS symbol versioning patches
From: Ulf Carlsson <ulfc@calypso.engr.sgi.com>
Date: Thu, 29 Jun 2000 18:20:43 -0700 (PDT)
Cc: "J. Scott Kasten" <jsk@tetracon-eng.net>, linux-mips@oss.sgi.com
In-reply-to: <20000630023352.E15960@bacchus.dhis.org>
References: <14671.21669.3126.181895@calypso.engr.sgi.com> <Pine.SGI.4.10.10006291446000.17956-100000@thor.tetracon-eng.net> <14683.40873.494836.164828@calypso.engr.sgi.com> <20000630023352.E15960@bacchus.dhis.org>
Sender: owner-linux-mips@oss.sgi.com
 > > Yes, it's just a matter of recompiling everything against it.  Keith
 > > Wesolowski is doing this.  We have found some minor problems with
 > > glibc that we haven't been able to resolve yet.  There is some bug in
 > > the dynamic linker, it tries to resolve symbols that aren't there in
 > > some packages.
 > > 
 > > There is also a problem with compiling gcc 2.96 natively, but I
 > > actually think that's a problem in gcc 2.96.  It shouldn't try to
 > > generate jump instructions like it does that in PIC code.
 > 
 > What jump instructions?  jal goes ok, the assembler knows how expand
 > them correctly for PIC code.

It uses j to perform branches that are out of range.  I think gcc
expects the assembler to translate it into a GOT16/LO16 thing if it's
out of range.

Ulf

<Prev in Thread] Current Thread [Next in Thread>