After successful configure build failed with:

make[3]: Entering directory
`/export/home/admin/xmms/not_tested/cairo-0.9.2/pixman/src'
if /bin/bash ../../libtool --tag=CC --mode=compile gcc -DHAVE_CONFIG_H -I. -I.
-I../.. -I../.. -I.     -O3 -mcpu=ultrasparc -mv8plus -mvis -funroll-loops -MT
ic.lo -MD -MP -MF ".deps/ic.Tpo" -c -o ic.lo ic.c; \
then mv -f ".deps/ic.Tpo" ".deps/ic.Plo"; else rm -f ".deps/ic.Tpo"; exit 1; fi
 gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../.. -I. -O3 -mcpu=ultrasparc -mv8plus
-mvis -funroll-loops -MT ic.lo -MD -MP -MF .deps/ic.Tpo -c ic.c  -fPIC -DPIC -o
.libs/ic.o
ic.c:151:3: warning: #warning "I havn't tested fbCompositeTrans_0888xnx0888() on
big endian yet!"
ic.c: In function `_cairo_pixman_composite':
ic.c:1686: warning: visibility attribute not supported in this configuration;
ignored
/usr/ccs/bin/as: "/var/tmp//ccQtauPI.s", line 4802: error: invalid operand
/usr/ccs/bin/as: "/var/tmp//ccQtauPI.s", line 4803: error: statement syntax
/usr/ccs/bin/as: "/var/tmp//ccQtauPI.s", line 4805: error: redefinition of
symbol "_cairo_pixman_composite"
/usr/ccs/bin/as: "/var/tmp//ccQtauPI.s", line 5636: error: statement syntax

Then I changed to  pixman/src and tried to compile it there

gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../.. -I. -O3 -MT ic.lo -MD -MP -MF
".deps/ic.Tpo" -c -o ic.lo ic.c ;
=>
/usr/ccs/bin/as: "ic.s", line 5065: error: invalid operand
/usr/ccs/bin/as: "ic.s", line 5066: error: statement syntax
/usr/ccs/bin/as: "ic.s", line 5068: error: redefinition of symbol
"_cairo_pixman_composite"
/usr/ccs/bin/as: "ic.s", line 5864: error: statement synta

.global _cairo_pixman_composite
        .type   _cairo_pixman_composite, #function
        .proc   020
_cairo_pixman_composite
...
.size   _cairo_pixman_composite, .-_cairo_pixman_composite
Now I commented out 
//slim_hidden_def(pixman_composite);
in ic.c and it compiled. Ok now the symbol is not hidden any more.
I think you can use mapfile to decide which modules should be hidden.
(I'll have a look for that later).
Maybe it's a problem of the sun assembler ...

Maybe the lib can be build somehow but as long it's not tested on
a big endian machine it could/should  not be used?