190794 - Build on Redhat Alpha Linux 7.1 fails.

Status: VERIFIED INVALID

(SeaMonkey :: Build Config, defect)

Description

[mach1]

User-Agent:       Mozilla/5.0 (X11; U; Linux alpha; en-US; rv:1.3b) Gecko/20030111
Build Identifier: Mozilla/5.0 (X11; U; Linux alpha; en-US; rv:1.3b) Gecko/20030111

A build of the browser attempted on 26-JAN-2003 fails. A previous attempted
build on 18-JAN-2003 failed in the same way. The last successful build was
performed on 11-JAN-2003. I will attach an extract of the log showing the
reported failure.

Reproducible: Always

Steps to Reproduce:
1.mkdir 26jan2003
2.cd 26jan2003
3.cvs co -f mozilla/client.mk
4.cd mozilla
5.gmake -f client.mk

My .mozconfig file contains the single line:
ac_add_options --enable-crypto
Actual Results:  
See attached extract of the log.

Expected Results:  
Completed the build successfully.

Comment 1

[mach1]

Attachment: Extract from the build log for 26-JAN-2003

Comment 2

[hacker formerly known as [e-mail]]

Given the time period, this may have been caused by the recombination of the
layout & content shared libs.  Do you see the same problem if you try doing a
--disable-debug build or --enable-debugger-info-modules=^layout,^content ?

Comment 3

[mach1]

I tried a build with each of the two options you mentioned with similar results.
I just added each option in turn to .mozconfig and re-issued the gmake command,
but I did not do a complete new download of the source from scratch. Both builds
failed on the build of libgklayout.so.

Comment 4

[hacker formerly known as [e-mail]]

Since those are configure flags, you would have to at least re-run configure for
them to take effect.  I would suggest doing a distclean first and building from
scratch with those flags.

Comment 5

[mach1]

I have attempted a new build from scratch with the additional options. The build
failed in the same manner as before.

Comment 6

[Andrew Schultz]

I tried building 1.3b on alpha (Compaq/Red Hat 7.2) and libgklayout.so linked fine.

Comment 7

[Andrew Schultz]

I did hit this bug trying to do a debug build with current CVS trunk, but adding
--disable-debug didn't help.  I'm trying to figure out what was different
between the 1.3b build and current trunk that broke it.

1.3b (RPM options, worked)
--prefix=/usr --libdir=/usr/lib --enable-optimize --disable-debug
--with-default-mozilla-five-home=/usrl/lib/mozilla-1.3b --enable-strip     
--disable-tests --disable-short-wchar --enable-nspr-autoconf --enable-crypto
--without-system-nspr --with-system-zlib --disable-xft --mandir=/usr/share/man

trunk (died trying to link libgklayout.so)
--disable-mailnews --disable-ldap --disable-freetype2 --disable-xprint
--disable-jsd --disable-postscript --disable-composer --disable-installer
--disable-mathml --disable-accessibility --disable-dtd-debug --disable-debug
--enable-strip

marking NEW

Comment 8

[Andrew Schultz]

adding --disable-tests --enable-optimize to the trunk build "fixed" it.  Is
there any reason that optimization would help here?

Comment 9

[John Morrison]

just as a general pointer, recombining layout and content libs will require
more swap space than before. mach1: how much swap space do you have allocated?

Comment 10

[Andrew Schultz]

the Alpha I'm compiling on has 256MB RAM and 512 MB swap.  I'll watch swap usage
next time it links (I think I noticed my i686 build get up to ~200MB total
memory usage).

Comment 11

[mach1]

My Alpha also has 256MB of RAM and a 534MB swap partition.

Comment 12

[Andrew Schultz]

my optimized debug build died just as the unoptimized non-debug build.  ld
occupied 80-90MB of memory and free swap space didn't drop below 440MB.

Comment 13

[Andrew Schultz]

I grabbed the most recent binutils release (2.13.2.1) and with that version,
libgklayout.so linked properly and Mozilla seems to work fine.
http://sources.redhat.com/binutils/

Red Hat 7.2's standard binutils is version 2.11.90.0.8-14

I filed a bug with Compaq/HP for this issue as it does not seem to be Mozilla's
fault:
http://alpha.crl.dec.com/bugzilla/show_bug.cgi?id=37

Comment 14

[Andrew Schultz]

FYI: the binutils-2.11.93.0.2-11 SRPM from Red Hat 7.3 / x86 rebuilt for Alpha
also works fine here

Comment 15

[hacker formerly known as [e-mail]]

Well, after some googling, I cannot find the abiword fix referenced by
http://alpha.crl.dec.com/bugzilla/show_bug.cgi?id=37 but
http://sources.redhat.com/ml/binutils/2002-11/msg00177.html seems to imply that
it's a binutils bug of some sort which matches Andrew's tests.  Marking invalid.

Comment 16

[Tracy Walker [:tracy]]

verified