X failure after userland upgrade

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

after upgrading my userland from several weeks old to
sources from  
yesterday I get relocation errors from XFree86's ELF
loader.
R_PPC_REL24 PLT offset too large 10fbac10.
Yup, that's more than 24bit. Any idea what could cause that
and how  
to fix it? Apparently AMD64 needed a fix there after
jemalloc was  
imported, could this be more fallout?

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBRxKAycpnzkX8Yg2nAQJEYQf/RpTIwfEINoA2yaOUWVmEZJ+TI1jC
C1Q/
m/4mPdT58banW6QKq4GOyOXYxREnRBbs10HSu5qDg1kg2R4bZh+AW0bW7ShR
yMvK
tKS+d7ZbDfDPFcCWDA0wFqBMxltmLH6ymGtwGdeANmtWata5WMQBFomWKh0w
BEyu
kjL1tuMChn4i453mtX4dx8OS4jHxJ0Grt3xCuOrTp2YqKI1/hQbio5k7S60A
s8sS
2Yn/gXTyW6CkPRPxtGIQmgbzJCFoJtfPHm0exgPTjkxPLI7jMo65a9rFH/Lv
+BOC
LhKBxvzx9S1TSKEIxjTlEZfC9nwCC2tCyDTFtU/gzmXe85kAeIxJrQ==
=dWZT
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Oct 14, 2007, at 16:49, Michael wrote:

> after upgrading my userland from several weeks old to
sources from  
> yesterday I get relocation errors from XFree86's ELF
loader.
> R_PPC_REL24 PLT offset too large 10fbac10.
> Yup, that's more than 24bit. Any idea what could cause
that and how  
> to fix it? Apparently AMD64 needed a fix there after
jemalloc was  
> imported, could this be more fallout?

Turns out it is jemalloc fallout - building a libc with  
USE_JEMALLOC=no makes X work again.
Now what side effects of the old malloc() is X relying on?

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBRxRo+spnzkX8Yg2nAQINAgf/UeMMVwsQkYXogpea1t3VmmtDssQW
fZhX
I7KKPtk3+SZ5ZH9e8+l2Z5wZWvujS85BzXppW1iRt2H+LwZKiPCR6ep7B7Iy
vC63
pnBslKHPaIIDbbmsYrsnfv6SDQJlVqwq4r0F9wzQsFBex2Y8wDFrrDQ9C8m7
mPDS
/75INTIIagbjIlSjdKgMhRT/nRXObv+fHvGiutu0tXUhESfbZyy6nGslFgy0
ck+Z
Vah6f3AfV7nTTStP4W4Yg/wEO7KGIFHW5DErUoqhGCHJB7l9C8Et8ztRm/1f
ssWG
aKROSv4AU31J31dcZyhncfqnVSE148X09w6SDePvGfqPvGNktBtNeQ==
=qGxg
-----END PGP SIGNATURE-----

On Tue, Oct 16, 2007 at 03:32:09AM -0400, Michael Lorenz
wrote:
> Turns out it is jemalloc fallout - building a libc with
 
> USE_JEMALLOC=no makes X work again.
> Now what side effects of the old malloc() is X relying
on?

The fact that the area between 1GB and 3GB was not used for
the heap.
The X11 server manually allocated memory with mmap() in that
area
to get memory below <4GB.

	Kind regards

-- 
Matthias Scheler                                  http://zhadum.org.uk/