Re: Emacs 27.1 released

["Mattias Engdegård" <mattiase@acm.org>]

12 aug. 2020 kl. 20.27 skrev Eli Zaretskii <eliz@gnu.org>:

> It has a potential of affecting a lot of code and customizations,
> since 'kbd' is used so widely.

I don't see how, but am not going to press the matter further.

>>>> 8ef84632c2 Accept lexical lambda in auto-insert-alist
>>> 
>>> I don't think this is important enough, but you could try convincing
>>> me by telling more details, including when this was introduced.
>> 
>> The code is old and assumes that a list not starting with 'lambda' cannot be a function. That ceased to be true when lexical binding was introduced; it was a test failure on master that alerted us to it. Fixing this obvious flaw just seemed natural and risk-free. It's nothing I'd fight for, in any case.
> 
> How easily would people bump into this in real-life use cases?  I
> think the answer to that will tell how important is it to fix it in
> Emacs 27.

We have no idea. Most bugs are never reported; people blame themselves and try a different approach, or just blame Emacs in general. That's why it's important to fix bugs we find even absent any concrete evidence that it has affected anyone else.

>>>> 60cd6cce55 Calc: GCD(0,x)=GCD(x,0)=|x|, not x (bug#41279)
>>> 
>>> I'm uneasy with this, as it would indirectly affect many functions.
>> 
>> Which many functions did you have in mind?
> 
> Its callers inside Calc.  Suppose there's some subtle issue with the
> change, or some unintended consequence.

Mind being more specific? I could certainly have overlooked something, but am not helped by these vague allegations in the slightest.

>  OTOH, how probable is it to have a negative x here?

The post-hoc probability is 1 because a user was affected by and reported the bug, but that's not the point: the erroneous simplification GCD(x, 0) -> x leads to trouble later on, as the report shows.