Re: MPS: hash tables / obarrays

[Helmut Eller]

On Mon, May 20 2024, Gerd Möllmann wrote:

> Helmut Eller <eller.helmut@gmail.com> writes:
>>    #2314# 635 #2194# 636 #2181# 637 #2022# 638 #1949# 639 #1891# 640 
>> #1640#)) #s(comp-data-container nil #s(hash-table test 
>>
>> So it could have something to do with hash tables after all.
>
> Could be, of course.

The hash table test wasn't scanned.  Here is a test to show it:

(progn
  (progn
    (defconst foo-sym (make-symbol "foo"))
    (define-hash-table-test foo-sym (function eql) (function sxhash-eql))
    (defconst foo-var (make-hash-table :test foo-sym)))
  (igc--collect)
  (message "%s" foo-var))

And with this it works:

diff --git a/src/igc.c b/src/igc.c
index fd206a0ff01..362c0cc32da 100644
--- a/src/igc.c
+++ b/src/igc.c
@@ -1518,6 +1518,11 @@ fix_hash_table (mps_ss_t ss, struct Lisp_Hash_Table *h)
     // FIXME: weak
     IGC_FIX12_NOBJS (ss, h->key, h->table_size);
     IGC_FIX12_NOBJS (ss, h->value, h->table_size);
+    const struct hash_table_test *p = h->test;
+    struct hash_table_test *test = (struct hash_table_test *)p;
+    IGC_FIX12_OBJ (ss, &test->user_hash_function);
+    IGC_FIX12_OBJ (ss, &test->user_cmp_function);
+    IGC_FIX12_OBJ (ss, &test->name);
   }
   MPS_SCAN_END (ss);
   return MPS_RES_OK;