Re: Segfault in read()

On Fri, Jul 12, 2024 at 10:25:15PM -0700, Ilya Zakharevich wrote:
> On Fri, Jul 12, 2024 at 07:25:14PM +0200, Bill Allombert wrote:
> > OK, but what is "pos" supposed to be  ?
> > 
> > What dbg_x(pos) gives ?
> 
> Oups, I remembered that there was something like dbg_x(), but after
> inspecting a couple of dbg_*() functions gave up — too early!  I
> append the results.
> 
> > For what I see, pos is a vector with 5 components, whose first component is a t_LIST 
> > which is corrupted.
> > 
> > So it is unclear if the bug is in writebin or read.
> 
> Now I it seems that I fixed most of MY bugs — only PARI’s shortcomings
> and limitations seem to be hit now (but this took about 250 millions
> tries of fuzzing!).  So I just attach the code (with one “workaround”
> commented out — to trigger the bug I was tracing through; it is
> irrelevant though to the segfault in question).
> 
> Load the code, then do
>   neg1=neg2=POS1=POS=pos2=[1..2] \\ buckets for debugging
>   fuzz_convhull3D_avoids0(19,[5],0x1,3,4)

Thanks, I have reported this bug as #2563.

Cheers,
Bill