On Tuesday, July 23rd, 2024 at 18:48, Gerd Möllmann <gerd.moellmann@gmail.com> wrote:
> Pip Cet pipcet@protonmail.com writes:
> 
> > On Tuesday, July 23rd, 2024 at 18:37, Eli Zaretskii eliz@gnu.org wrote:
> > 
> > > So this is IGC_OBJ_STRING_DATA, but why doesn't igc_dump_finish_obj do
> > > its job in this case? What are we missing?
> > 
> > I'm pretty sure it's the bignum thing, currently testing a fix. I
> > underestimated the complexity of how bignums are dumped, and wrongly
> > assumed that on 64-bit systems, we would be dumping bignums; it seems
> > we don't do so anymore. (BIGNUM_DATA, unlike the bignum PVEC, doesn't
> > have a header; it should, and if I add bignums to the dump manually
> > things work now...).
> > 
> > Pip
> 
> 
> Ok, good to hear :-).

Here's a patch which includes both a reproducer for the bug (defining a global bignum in loadup.el which then gets dumped) and a tentative fix.

If bignum dumping is indeed broken on all architectures, including just the change to loadup.el should reproduce the bug even on 64-bit archs.

Pip