Ludovic Courtès <ludo@gnu.org> skribis:

> I’ll try to combine that with incremental marking of the weak table, but
> I’m not very hopeful.

Here’s an attempt to mark tables incrementally.

Unfortunately it misses references sometimes (e.g., values of a weak-key
table do not get marked even though they should.)

I’m unsure whether incremental marking is actually doable: the mark
procedure only knows its own part of the mark state, not the global mark
state.  For instance, it cannot distinguish between different mark
phases.  I tried to fix that by memorizing the ‘GC_mark_no’ value we
were in when we left the mark procedure, but that didn’t help.

Ideas?

Ludo’.