bug#56623: memory leak introduced by new nonrecursive lisp reader

[Gerd Möllmann]

Gerd Möllmann <gerd.moellmann@gmail.com> writes:

> Mattias Engdegård <mattiase@acm.org> writes:
>
>> If either makes things better, maybe we can come up with a solution
>> that doesn't hurt performance.
>
> This might be completely off, but anyway:
>
> lread.c uses free_cons.  Is it certain that the cons cell is never
> marked at that point?  Not that I see it happening, or could explain how
> exactly GC would run while that code executes.  Maybe it's worth adding
> a check in free_cons?

One candidate: A hash table with user-defined :test.  Invoking the
user-defined functions can GC.