Re: master 2de25accaf 4/4: Warn about `condition-case' with quoted condition names

Re: master 2de25accaf 4/4: Warn about `condition-case' with quoted condition names

[Stefan Monnier]

> +        (when (and (eq (car-safe condition) 'quote)
> +                   (cdr condition) (null (cddr condition)))
> +          (byte-compile-warn-x
> +           condition "`condition-case' condition should not be quoted: %S"
> +           condition))

Shouldn't we instead warn about any unknown error signals?


        Stefan

Re: master 2de25accaf 4/4: Warn about `condition-case' with quoted condition names

[Mattias Engdegård]

29 dec. 2022 kl. 15.45 skrev Stefan Monnier <monnier@iro.umontreal.ca>:

> Shouldn't we instead warn about any unknown error signals?

Ideally both -- special-casing quoted arguments gives better diagnostics.
But do we know all errors statically at compile-time?

Re: master 2de25accaf 4/4: Warn about `condition-case' with quoted condition names

[Stefan Monnier]

>> Shouldn't we instead warn about any unknown error signals?
> Ideally both -- special-casing quoted arguments gives better diagnostics.
> But do we know all errors statically at compile-time?

Error identifiers should have an `error-conditions` property.


        Stefan

Re: master 2de25accaf 4/4: Warn about `condition-case' with quoted condition names

[Mattias Engdegård]

29 dec. 2022 kl. 16.33 skrev Stefan Monnier <monnier@iro.umontreal.ca>:

>> But do we know all errors statically at compile-time?
> 
> Error identifiers should have an `error-conditions` property.

You left the `error-conditions` property check commented out in `byte-compile-condition-case` so maybe it wasn't quite that straightforward?

Re: master 2de25accaf 4/4: Warn about `condition-case' with quoted condition names

[Stefan Monnier]

Mattias Engdegård [2022-12-29 16:50:37] wrote:
> 29 dec. 2022 kl. 16.33 skrev Stefan Monnier <monnier@iro.umontreal.ca>:
>>> But do we know all errors statically at compile-time?
>> 
>> Error identifiers should have an `error-conditions` property.
>
> You left the `error-conditions` property check commented out in
> `byte-compile-condition-case` so maybe it wasn't quite that straightforward?

I can't remember what came up, but my guess is that it's a question
of making the byte-compiler recognize `define-error` and things like
that (i.e. try and reduce the number of false positives where the error
is unknown at compile time but will be known at run-time).


        Stefan