Nikodemus Siivola <nikodemus@...> writes:
> As of 188.8.131.52 (below for convenience) WITH-LOCKED-HASH-TABLE also
> disables interrupts for its body.
>>From the commit message it is not clear to me why.
> WITH-LOCKED-HASH-TABLE should IMO be considered a user lock. To me it
> is morally analogous to a user have an external lock and grabbing that
> around accesses to a hash-table.
> I'm not sure if the meaning here is that SBCL's internal hash-tables
> should keep interrupts disabled -- which I can certainly understand --
I think this was the intention.
> or if there is something non-obvious in the hash-table code that makes
> interrupts arriving during a locked section fundamentally unsafe?
It doesn't tickle my memory, and I almost surely would have mentioned
such a thing in the commit message, had what you are asking about been
found to be the case.