[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: GCD bug?
From: |
Ken Kahn |
Subject: |
Re: GCD bug? |
Date: |
Mon, 7 Apr 2003 16:32:50 -0700 |
Sorry about the delay.
ASSERT (usize * GMP_NUMB_BITS >= d);
aborts in bdivmod.c on line 64. usize is 69 and d is 4294967269 (both
decimal).
I'll substitue the C version of submul_1 and report soon.
Thanks for your help with this.
Best,
-ken
----- Original Message -----
From: "Kevin Ryde" <address@hidden>
To: "Ken Kahn" <address@hidden>
Cc: <address@hidden>
Sent: Tuesday, April 01, 2003 3:10 PM
Subject: Re: GCD bug?
> "Ken Kahn" <address@hidden> writes:
> >
> > that it crashes when usize=-1
>
> Well that's definitely wrong, but it's not obvious how that would
> arise. The gcd loop is supposed to maintain usize>=vsize, and to go
> to the binary finish-up when vsize<=2.
>
> Some of the assertions might help narrow it down. Edit or add to
> config.h,
>
> #define WANT_ASSERT 1
>
> and recompile everything. See if you can get the C versions of the
> low level routines too, not the assembler versions, since the latter
> don't have checking. (Should be as easy as compiling submul_1.c
> instead of submul_1.asm.)
>
> What you'll hope to see is some assertion go off at a nice early
> stage. It'll do an abort(), which you can catch in the debugger and
> go from there.
>
> --
> All followups to address@hidden please.
>
- Re: GCD bug?, Kevin Ryde, 2003/04/01
- Re: GCD bug?,
Ken Kahn <=