bug#46670: 28.0.50; [feature/native-comp] possible miscompilation affecting lsp-mode

[Andrea Corallo via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

Pip Cet <pipcet@gmail.com> writes:

> On Mon, Feb 22, 2021 at 11:23 AM Andrea Corallo <akrl@sdf.org> wrote:
>> Pip Cet <pipcet@gmail.com> writes:
>> > On Mon, Feb 22, 2021 at 10:04 AM Pip Cet <pipcet@gmail.com> wrote:
>> > (Except for this one:
>> >
>> >       ((defun comp-tests-ret-type-spec-f (x)
>> >          (unless (symbolp x)
>> >            x))
>> >        (not symbol))
>> >
>> > If I'm reading that correctly, it tests that (unless (symbolp x) x)
>> > isn't a symbol, which it usually is)
>>
>> Yep, it verifies that this function has as inferred return type (not
>> symbol).
>
> Which means the return value shouldn't ever be a symbol, right?
> Because it's nil, which is a symbol, when (symbolp x). Am I missing
> something here?

Sorry I though the question was on the test mechanism and wasn't pay
attention to the specific testcase content :/

Right that's clearly a bug in `comp-cstr-union-1-no-mem' that was
missing to check that no negative type is shadowing any positive type
coming from values and giving-up returning t in case).

Good catch thanks! :) Should be fixed by d6227f6edc.

  Andrea

PS as I see you are interested into this part of the compiler, I find
typically handy to exercise this logic with like:

(let ((comp-ctxt (make-comp-cstr-ctxt)))
  (comp-cstr-to-type-spec
   (comp-type-spec-to-cstr '(or (not symbol) null))))

We'll probably see other bugs in this area cause is tricky, is important
we build the best coverage we can for this in the testsuite.