Re: Redisplay hook error backtraces

[Alan Mackenzie <acm@muc.de>]

Hello, Stefan.

On Thu, Jul 14, 2022 at 12:18:51 -0400, Stefan Monnier wrote:
> >> Aren't they all go through safe_call?

> > They do, yes, but so do other things that we don't want to engage the
> > backtrace mechanism for.

> Actually, I'm not sure that's the case.  Basically the question is
> whether `safe_call` is used because we don't care about any errors that
> might be thrown (akin to ignore-errors`) or whether we do care but don't
> want to let them escape and ruin our lunch (akin to
> `with-demoted-errors`).

> Being in redisplay is usually a good reason why we don't want to let
> errors escape, even we do care about those errors (and hence we need
> to use a mechanism like the one you're implementing in order to debug
> those problems), but there can be other such cases that also use
> `safe_call` and where we would also want to use your new code.

There are approximately 38 calls to something like safe_call in the C
sources, about half of which are in xdisp.c.  That's a lot of scope for
irritation if non-events continually swamp the real errors in redisplay's
Lisp hooks.

It feels like we would want some mechanism to filter out "some" safe_call
calls, and things are getting complicated fast.  The scope of the
exercise has already expanded considerably, from just
fontification-functions to all redisplay hooks.

>         Stefan

-- 
Alan Mackenzie (Nuremberg, Germany).