Re: MPS: hash tables / obarrays

[Gerd Möllmann]

Helmut Eller <eller.helmut@gmail.com> writes:

> On Sun, May 19 2024, Gerd Möllmann wrote:
>
>> If someone could check the obarray related changes I made, that would be
>> nice.
>
> Does this only work if the dump is registered as exact root or also with
> ambiguous roots?

I only works with exact roots ATM. We could support both by changing
igc_make_hash_table_vec to add a root for the vector for hash tables
that are in the dump. So, the same that was in hash_table_alloc_kv
before.

>
> Because for me, comp-tests now fails reliably.  First I thought this has
> something to do with the dump being an exact root, but when I tried with
> ambiguous roots and discovered that comp-tests fails even faster.
> Basically immediately, while with exact roots it takes 2-3 minutes.

Alas, I can't currently configure with native-comp because something in
the libgccjit I got from Homebrew changed. It can't find libgcc_s.dylib
anymore.

What does it say when it fails?