Hi Chris,

Thanks for your support.  :-)

Christopher Allan Webber <cwebber@dustycloud.org> skribis:

> +      else
> +        {
> +          /* The move to BDW-GC with Guile 2.0 introduced some bugs
> +             related to weak hash tables, threads, memory usage, and the
> +             alloc lock.  We were unable to fix these issues
> +             satisfactorily in 2.0 but have addressed them via a rewrite
> +             in 2.2.  If you see this message often, you probably want
> +             to upgrade to 2.2.  */
> +          fprintf (stderr, "guile: warning: weak hash table corruption "
> +                   "(https://bugs.gnu.org/19180)");
> +          len = 0;
> +        }
>
> Guess reverting this patch means this comment also should be amended!

Indeed.  :-)

The attached patch fixes the bug for me.  Having spend days on
weak-table stuff, the bug looked (described in the commit log below)
almost obvious to me.

This tight loop used to trigger the bug after a few seconds; it no
longer does after several minutes:

  (define table
    (make-weak-key-hash-table))

  (let loop ((i 0))
    (unless #f
      (hashq-set! table (make-list 1000) i)
      (loop (1+ i))))

Also, it no longer displays the pathological behavior shown in
<https://bugs.gnu.org/28590>.

Of course, even better if people could test the two patches and confirm
that it works for them.

Then if there are no objections I’d like to merge them in ‘stable-2.2’.

Ludo’.