Re: Debugging a XS problem.. where to start?

On 16/01/12 21:28, Marvin Humphrey wrote:
> On Mon, Jan 16, 2012 at 09:11:27PM +0000, Alberto Simões wrote:
>> I do not use valgrind for some time, but this seems awkward:
>>
>> [ambs@stravinski FreeLing-old]$ valgrind -- /usr/bin/perl -I blib/arch/
>> -I blib/lib/ t/01-bindings/01-tokenizer.t
>> ==13831== Memcheck, a memory error detector
>> ==13831== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
>> ==13831== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
>> ==13831== Command: /usr/bin/perl -I blib/arch/ -I blib/lib/
>> t/01-bindings/01-tokenizer.t
>> ==13831==
>>
>>
>> and... nothing... I need to kill it.
>> Ideas?
>
> No memory errors have occurred yet in your program.
>
> IMO, you have given up too soon before reporting back.  I would advocate the
> use of a standard debugging technique to see where things are getting stuck:
> try sprinkling "warn" statements into your test file to see how far you get.
>
> Please bear in mind that running under Valgrind is typically 40x slower, so
> heavy tests can take a long time.

Found out that the test fails when the XS code calls a method in the 
library (in fact, tries to create an object of a specific type).

Now debugging inside that library will be almost impossible :-/
let me try again valgrind and gdb...

thanks