llvm dev - Feb 2010 - [LLVMdev] using bugpoint to find miscompilation/code generation problems?

Sorry to ask about a documented use of bugpoint, but I am having trouble
understanding what to do.
 
I would like to be able to use bugpoint to simplify test cases, when the
symptom is a run time difference in the testprogram behavior. That looks
like what is described in the documentation as "miscompilation or
perhaps code generation problems". There are examples in the LLVM doc
showing runs with -output argument to supply the expected output, -args
to give the program arguments, and -tool-args give arguments to, in this
case, llc. The problem is that I can't seem to get bugpoint to run this
way, and its probably my fault.

Can someone give me an example for LLVM 4.2.1 that runs bugpoint in this
way:  
 
1) I want to run a test compiled to bitcode (this is the bugpoint
input). I have this.
2) The bitcode compiled with llc to a .s file by bugpoint.
3) Then the .s file linked with gcc by bugpoint.
4) The final file run automatically by bugpoint against oggenc output
compiled by llvm-gcc straight to native code
 
It then iterates until a simplified test case is created whose output is
still different from that of the native code.
 
thanks,
-David
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20100216/9d895b3c/attachment.html>

Hi David,
> It then iterates until a simplified test case is created whose output is 
> still different from that of the native code.
isn't that exactly what you want: a small test-case that shows the problem?

Ciao,

Duncan.

Yes, its what I want, but I can't figure out how to get bugpoint to
accomplish this.
 
-David

________________________________

From: Duncan Sands [mailto:baldrick at free.fr]
Sent: Thu 2/18/2010 6:06 AM
To: David Dunkle
Cc: llvmdev at cs.uiuc.edu
Subject: Re: [LLVMdev] using bugpoint to find miscompilation/code generation
problems?



Hi David,
> It then iterates until a simplified test case is created whose output is
> still different from that of the native code.
isn't that exactly what you want: a small test-case that shows the problem?

Ciao,

Duncan.


-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20100218/6cc05c3a/attachment.html>