On 02/09/2011 09:13, Erik de Castro Lopo wrote:
Hi all,
For some time I have been working in short bursts on debugging trac
bug #5111 which actually turns out to be a manifestation of bug
#2972.
Reading bug #2972 shows that at different times, the segfault was
occurring in different places at different times (ie the curses
library, libedit and now somewhere else for me).
This makes me wonder if I am actually running into a PowerPC
architecture problem where relative jumps are limited to addresses
that can be expressed in 24/26 bits (24 actual bits shifted up
two bits because instructions are 4 byte aligned).
Anyone have any suggestions on how to progress this further?
I don't know anything about PPC, but I might be able to give you general
help with debugging the linker. To start with you want to compile stage
2 with debugging turned on:
$ cd ghc
$ rm stage2/build/tmp/ghc-stage2
$ make 2 GhcDebugged=YES
then run it with +RTS -Dl to get linker debug output (direct the output
to a file, there will be lots of it).
You probably also want to run GHCi under gdb to see where the crash is
happening, so check out
http://hackage.haskell.org/trac/ghc/wiki/Debugging/CompiledCode
Make sure you have an architecture manual, an ABI specification, and
some ELF documentation to hand.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
Cvs-ghc@haskell.org
http://www.haskell.org/mailman/listinfo/cvs-ghc
