glibc and rpm

[Jakub Bogusz]

Please check if ld.so is _REALLY_ not stripped now - for me those rpm
macros hacks didn't work - because LDFLAGS="-s" when building glibc and
linker strips ld.so *during build*!

To avoid this, we can set LD_FLAGS to something, eg. " ", and change
macros.pld (SOURCES/rpm.macros in repo) to not set LDFLAGS if it's set
already:
- LDFLAGS="${LDFLAGS}%{!?debug: -s}" ; export LDFLAGS ; \
+ LDFLAGS="${LDFLAGS:-%{!?debug: -s}}" ; export LDFLAGS ; \

Then we may strip ld.so in glibc.spec by:
%{!?debug:strip -g -R .comment -R .note $RPM_BUILD_ROOT/lib/ld-%{version}.so}

(BTW: rpm-brp-strip.patch does nothing. brp-strip doesn't even strip
files identified by file(1) as "shared object", and ld.so is shared
object. Also, we don't use brp-* - you can check this by:
rpm --eval '%__spec_install_post' )


Second problem is buggy rpm's installplatform script.
I wasn't even able to build rpm 4.0 without patching it (on i586)
and reported this on pld-list on Dec 25.
Because noone replied, I thought this was only my problem.
But please look at rpm-4.0*.i[356]86.rpm packages on ftp, in
/usr/lib/rpm/i[3456]86-pld-linux directories. We have optflags
"-O2 -march=i686" for i586 target. (but it seems that (for now?)
optflags only from rpmrc file are used, not from these directories,
so it didn't break anything).


I'm going to put above changes to repo now.
I hope I won't break anything... For me - it works.


BTW: what do you think about adding libtool to Requires list in
rpm-build? libtool is required to build most autoconf-based shared
libraries.


OT: something is wrong with SPECS module in anoncvs - on each
"cvs up SPECS" I get:
cvs server: nothing known about SPECS/perl-CDB_File.spec

-- 
Jakub Bogusz
http://prioris.mini.pw.edu.pl/~qboosh/