Khem Raj wrote:
On Sat, Oct 2, 2010 at 4:39 AM, Holger Freyther <[email protected]> wrote:
Hi all,
today I wanted to make the generated toolchain free of RPATH and I was a bit
surprised to see that even the target libraries contain the rpath and that it
is not caught by insane.bbclass.
I have added a more strict check to insane.bbclass to uncover the issue and I
am fighting neon and its manual libtool usage right now... it might be just
easier to use chrpath to fix the resulting libraries.
any opinion?
I think we should try to get the rpaths right without chrpath. so your
insane patch
seems ok. May be to avoid lot of packages failing this could be turned
on as a warning
and once we fix the packages failing this then we could turn it into error.
Initial link time $ORIGIN being correct is a lot more complex than being
wrong and fixing it up. I got it being correct for the non-autoconf
case and then made autotools.bbclass also get it right initially and
then patched gcc/gdb/binutils (which are the difficult ones that run
configure again) but it's really much more work. I'll try and RFC the
patches on Monday, if I've still got the branch around.
--
Tom Rini
Mentor Graphics Corporation
_______________________________________________
Openembedded-devel mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
