[LLVMbugs] opt segfaults

[Chris Lattner]

On Sat, 17 Apr 2004, Anthony Evans wrote:

> I am using LLVM to compile some programs, and opt crashes reproducibly on
> some input. Here is the tail of the bugpoint output:

Hrm, the -cee pass is not a stable pass (it is documented as such in the
release notes).  Is this blocking you?  You should just be able to not use
it...

> *** Attempting to perform final cleanups: Instruction does not dominate
> all uses!
>          %tmp.171.i = seteq sbyte %tmp.9, 0              ; <bool> [#uses=3]
>          %tmp.171.i.fw_frontier = phi bool [ %tmp.171.i,
> %shortcirc_next.3.i.endif.11.i_crit_edge ], [ %tmp.171.i.fw_merg
> e1, %endif.11.i.then.13.i_crit_edge ]           ; <bool> [#uses=0]
> Broken module found, compilation aborted!
> bugpoint[0x8365ae9]
> /lib/libc.so.6(abort+0x1d7)[0x4012ce27]
> bugpoint((anonymous namespace)::Verifier::abortIfBroken()+0x10a)[0x835848a]
> Crashed with signal #6
> Emitted bytecode to 'bugpoint-reduced-simplified.bc'
>
> *** You can reproduce the problem with: opt bugpoint-reduced-simplified.bc
> -cee
>
> This can be reproduced in 1.2 and in the current CVS (HEAD).
>
> This is on Linux 2.6.5 (+ Andrew Morton's mm-6 patchset), glibc 2.3.3
> (+ NPTL patches) on an AMD Barton 3000+ with 2GB RAM.
>
> gcc 3.3 and 3.4 (current CVS) and intel icc 8.0 can compile and optimise
> the code without problems on the same system.
>
>
> Tony

-Chris

-- 
http://llvm.cs.uiuc.edu/
http://www.nondot.org/~sabre/Projects/