bug#38912: 27.0.60; PDumper meets segmentation fault when evil is loaded

[Pip Cet <pipcet@gmail.com>]

On Sun, Jan 5, 2020 at 6:46 PM Eli Zaretskii <eliz@gnu.org> wrote:
> In my debug build of Emacs 27.0.60 I get an assertion violation while
> dumping, which probably is already a sign of trouble.

I think that's unrelated, but here's some analysis anyway, in case we
want to fix this bug:

Different bytecode objects may be `equal', but have different sxhashes.

(equal #[0 "" [] 0]
       #[0 "" [] 0])
(sxhash #[0 "" [] 0])
(sxhash #[0 "" [] 0])

When such bytecodes are used as keys for a hash table using
hashfn_equal, I believe the result is, fairly obviously, two hash
table entries for equal keys; when the hash table is rehashed, we may
be unlucky enough to retrieve the wrong one, leading to the crash.

That's what cl-generic.el does for the table cl--generic-dispatchers,
which I believe is what you were looking at.

>   #2  0x01326817 in check_hash_table_rehash (table_orig=XIL(0xa000000006288090))
>       at pdumper.c:2684

Can you confirm table_orig is cl--generic-dispatchers?

Again, I doubt this is related to the original bug. That `equal'
behaves weirdly like this is a problem I've mentioned before (when
(equal a b) signals but (equal b a) succeeds), but the consensus then
was not to change it, so it's possible this is only the rehashing
check needing to be more careful.