Re: [Bug-gsl] UNEXPECTED BUG in gsl_sf_lngamma_sgn_e !

[Brian Gough]

At Thu, 31 May 2007 23:21:43 +0200,
Nenad Sakan wrote:
> First of all, I hav compiled gsl 1.9 on gnu linux (slackware) x86
> processor.  By some mistake the debugger had pointed me to the
> previous value of the epsilon (not the real where it calls the
> lngama with x = 0). I have started to print out each parameter that
> goes to the function and noticed that realy it does not break for x
> != 0, what is expected).
> 
> I have tested it after that with a simple routine where I have
> passed the value that GDB pointed out and noticed that all goes well
> (that is also expected). For some reason backtrace in GDB has
> printed the value for 1F1 and further on which is not the step at
> which error occured but one before!?!?

Hello,

If the code is compiled with optimisation the values shown in the
debugger can be different, as some temporary results are only held in
registers and not stored back to the variables.  

( http://www.network-theory.co.uk/docs/gccintro/gccintro_51.html )

To see the true values the code and library need to be compiled
without optimisation (-g -O0).

-- 
best regards,

Brian Gough

Network Theory Ltd,
Publishing Free Software Manuals --- http://www.network-theory.co.uk/