X-pop3-spooler: POP3MAIL 2.1.0 b 4 980420 -bs-
Date: Thu, 10 Sep 1998 16:50:14 -0500 (EST)
From: Steven Snyder <ssnyder AT indy DOT net>
X-Sender: ssnyder AT indy1
To: pgcc mailing list <beastium-list AT Desk DOT nl>
Subject: Anyone build glibc v2.0.7 with pgcc v1.1a?
Message-ID: <Pine.SUN.3.96.980910164850.14173A-100000@indy1>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: Marc Lehmann <pcg AT goof DOT com>
Status: RO
X-Status: A
Content-Length: 1397
Lines: 28


I'm attempting to build (and test!) glibc with pgcc but am encountering
problems that are not seen when compiling with gcc.  The build itself
shows no obvious problems, but the self test ("make check") will not
complete if pgcc is used to compile the code.

I installed the source package (glibc-2.0.7-19.src.rpm) and installed it
on my RH 5.1 system.  I can successfully build the binaries (using the
command "rpm -bc /usr/src/redhat/SPECS/glibc-2.0.7.spec") with either
compiler.  After the binaries are generated, I go to the build directory
and run "make check".  If the packages was built with gcc v2.0.7.3 the
test completes without incident.  If the package was built with pgcc
v1.1a, the test is halted when an error occurres in attempting to execute
linuxthreads/ex1.

The error is the inability to resolve the symbol "__register_frame_info"
when loading some shared library.  Dumping the symbol map for (wannabe)
executable ex1, I see that the missing symbol is referenced in csu/crti.o.
(BTW, I am building with the 2.0.7.3 libraries already installed via the
glibc-2.0.7.3-19.i386.rpm package.)

This is driving me crazy.  In both builds (gcc & pgcc) the exact same
switches are used: "-O2 -m486 -fno-strength-reduce -g -DNDEBUG=1".  Can
anyone suggest a reason for the missing symbol?  I'm sure not going to
install these libs if they can't even pass a set of known tests.

Thanks.