[Qemu-devel] Re: ARM page crossing inside insn? (Re: ARM ethernet fixes)

[Antti P Miettinen <ananaza@iki.fi>]

Antti P Miettinen <ananaza@iki.fi> writes:
> Something like this?

Hmm.. I suppose the branch should be the last op to generate if PC is
loaded by the ldm. So the possible gen_bx should be after the
writeback? But what about the S bit? Or is this all highly academic?
Does ldm ever load pc in practice?

Anyway - I managed to rebuild firefox with debug info and found where
it segfaulted. And with the below change it runs :-)

diff -ur firefox-1.4.99+1.5rc3.dfsg/xpcom/reflect/xptcall/src/md/unix/xptcinvoke_arm.cpp /usr/local/arm/sid/home/apm/src/firefox-1.4.99+1.5rc3.dfsg/xpcom/reflect/xptcall/src/md/unix/xptcinvoke_arm.cpp
--- firefox-1.4.99+1.5rc3.dfsg/xpcom/reflect/xptcall/src/md/unix/xptcinvoke_arm.cpp	2005-10-14 00:33:10.000000000 +0300
+++ /usr/local/arm/sid/home/apm/src/firefox-1.4.99+1.5rc3.dfsg/xpcom/reflect/xptcall/src/md/unix/xptcinvoke_arm.cpp	2005-12-25 15:15:46.000000000 +0200
@@ -212,7 +212,7 @@
     "add	sp, sp, r4	\n\t"	/* restore stack pointer		*/
     "mov	%0, r0		\n\t"	/* the result...			*/
     : "=r" (result)
-    : "r" (&my_params)
+    : "r" (&my_params), "m" (my_params)
     : "r0", "r1", "r2", "r3", "r4", "ip", "lr", "sp"
     );
     
-- 
http://www.iki.fi/~ananaza/