--==_Exmh_27409472P
Content-Type: text/plain; charset=us-ascii

In message <20030325205704.GA10832@antioche.lip6.fr>, you write: 

-> Hi,
-> OK, I think I tracked down the problem of perl not building properly
-> on my sgimips. Background: I'm seeing a different behavior of the build 
-> process on i386 and sgimips. Note that I'm running current on the sgimips.

This sounds like the same ld_elf.so lossage that affects tcsh and bash as
well (both of which define their own free/malloc -- the C library ends up
using it's own malloc but then calls the application's free(), leading to
a spew of messages from the shell about invalid blocks being freed.

I can't check right now, but I know there are PR's against for for both
bash and tcsh (though I'm not sure if against port-sgimips, lib/bin or
generic port-mips)

I spent a little time looking at it, but didn't make it very far; at
least digging through the LD_DEBUG output from ld_elf.so didn't lead
to an immediate clue -- I think it appeared that some instances of
the symbol were fixed up at ld_elf.so invocation time whereas others
were done via lazy fixup and that may have been why some got it right
while others didn't, but I can't remember for sure.

--rafal

----
Rafal Boni                                              rafal.boni@eDial.com
 PGP key C7D3024C, print EA49 160D F5E4 C46A 9E91  524E 11E0 7133 C7D3 024C


--==_Exmh_27409472P
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Exmh version 2.4 06/23/2000

iD8DBQE+gMqQEeBxM8fTAkwRAuavAJ9S3UrZ6v3yEhehhLE7BFQjRiqVYACeKE63
j48tOrwEHlGiBdMso7nmSbU=
=AU+E
-----END PGP SIGNATURE-----

--==_Exmh_27409472P--