bug#68244: hash-table improvements

[Andy Moreton <andrewjmoreton@gmail.com>]

On Mon 15 Jan 2024, Mattias Engdegård wrote:

> 15 jan. 2024 kl. 13.31 skrev Eli Zaretskii <eliz@gnu.org>:
>
>>> Bisect identifies this as the bad commit:
>>>  d3cefd3e9835 ("Leaner hash table dumping and thawing")
>
> Thank you, I can reproduce the (or at least an) assertion failure in this
> commit. However it's probably a ghost error: although I've tried to make sure
> that there is a correct build at each point, there have been a lot of work
> going on rearranging things for correctness and clarity, and apparently I
> never built this particular one with checking enabled the last time.
>
> However this assertion should be removed by the next commit so the error is
> very short-lived (and it's just an over-zealous assertion, not an actual
> flaw). What problems do you observe when building master?
>
> (And yes, I should update some hashes in pdumper.c; it's been a while.)

I bisected from newer master to get to that point. Bootstrapping master
commit f19f5604deb7 ("Update pdumper hashes for buffer and Lisp_Hash_Table")
is still failing, and trying to continue the buid by retrying after that
is glacially slow, so something is badly broken at that point.

Note that the same commit bootstrapped after "git clean -xdf" still has
multiple backtraces when building without ENABLE_CHECKING.

    AndyM