GNU bug report logs - #76949
30.1; --debug-init and condition-case-unless-debug

Previous Next

To reply to this bug, email your comments to 76949 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: "Nussbaum  Ferdinand" <ferdinand.nussbaum <at> inf.ethz.ch>
To: "bug-gnu-emacs <at> gnu.org" <bug-gnu-emacs <at> gnu.org>
Cc: "me <at> bcc32.com" <me <at> bcc32.com>,
 "monnier <at> iro.umontreal.ca" <monnier <at> iro.umontreal.ca>
Subject: 30.1; --debug-init and condition-case-unless-debug
Date: Tue, 11 Mar 2025 17:13:56 +0000

[Message part 1 (text/plain, inline)]

Errors inside condition-case-unless-debug do not trigger the debugger when the
--debug-init startup option is set (and debug-on-error is not additionally and
explicitly set to a non-nil value).

This can be reproduced, for example, by adding the following to init.el:

(condition-case-unless-debug nil
    (error "error")
  (error
   (message "error caught")))

and restarting Emacs with --debug-init. In Emacs 29 this enters the debugger
(as expected), while in Emacs 30 the message "error caught" is displayed.

On a related note, the Emacs Lisp reference manual still says the following: "[...]
use the option ‘--debug-init’. This binds debug-on-error to t while loading the
init file, [...]".
(https://www.gnu.org/software/emacs/manual/html_node/elisp/Error-Debugging.html#Error-Debugging)
AFAICT this is no longer correct since the implementation uses handler-bind instead.

(Originally reported in https://github.com/syl20bnr/spacemacs/issues/16871
and https://github.com/syl20bnr/spacemacs/issues/16619#issuecomment-2708510954.)

Best,
Ferdinand

[Message part 2 (text/html, inline)]

Message #8 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: "Nussbaum Ferdinand" <ferdinand.nussbaum <at> inf.ethz.ch>
Cc: "bug-gnu-emacs <at> gnu.org" <bug-gnu-emacs <at> gnu.org>,
 "me <at> bcc32.com" <me <at> bcc32.com>
Subject: Re: 30.1; --debug-init and condition-case-unless-debug
Date: Tue, 11 Mar 2025 16:41:42 -0400

> Errors inside condition-case-unless-debug do not trigger the debugger when the
> --debug-init startup option is set (and debug-on-error is not additionally and
> explicitly set to a non-nil value).

Indeed this is the direct consequence of:

> AFAICT this is no longer correct since the implementation uses handler-bind instead.

It seems cumbersome to fix this without re-introducing the problem that
`handler-bind` addresses, so I'm in favor of fixing the doc rather than
the code.


        Stefan

Message #11 received at 76949 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Stefan Monnier <monnier <at> iro.umontreal.ca>
Cc: ferdinand.nussbaum <at> inf.ethz.ch, me <at> bcc32.com, 76949 <at> debbugs.gnu.org
Subject: Re: bug#76949: 30.1; --debug-init and condition-case-unless-debug
Date: Wed, 12 Mar 2025 14:55:43 +0200

> Cc: 76949 <at> debbugs.gnu.org, me <at> bcc32.com
> Date: Tue, 11 Mar 2025 16:41:42 -0400
> From:  Stefan Monnier via "Bug reports for GNU Emacs,
>  the Swiss army knife of text editors" <bug-gnu-emacs <at> gnu.org>
> 
> > Errors inside condition-case-unless-debug do not trigger the debugger when the
> > --debug-init startup option is set (and debug-on-error is not additionally and
> > explicitly set to a non-nil value).
> 
> Indeed this is the direct consequence of:
> 
> > AFAICT this is no longer correct since the implementation uses handler-bind instead.
> 
> It seems cumbersome to fix this without re-introducing the problem that
> `handler-bind` addresses, so I'm in favor of fixing the doc rather than
> the code.

Feel free to fix the documentation.  But it would be nice if we could
tell users what to use instead of condition-case-unless-debug in these
cases.

Message #14 received at 76949 <at> debbugs.gnu.org (full text, mbox):

From: Aaron Zeng <me <at> bcc32.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: ferdinand.nussbaum <at> inf.ethz.ch, me <at> bcc32.com,
 Stefan Monnier <monnier <at> iro.umontreal.ca>, 76949 <at> debbugs.gnu.org
Subject: Re: bug#76949: 30.1; --debug-init and condition-case-unless-debug
Date: Wed, 12 Mar 2025 22:24:07 -0400

With apologies to the maintainers, I accidentally replied directly 
to
Stefan rather than reply-all.  Copying in some context from our
exchange below (hand-edited from the most recent email I received 
from
Stefan):

> Aaron Zeng <me <at> bcc32.com> writes:
>
> > Stefan Monnier <monnier <at> iro.umontreal.ca> writes:
> >
> > >> It seems cumbersome to fix this without re-introducing the 
> > >> problem that
> > >> `handler-bind` addresses, so I'm in favor of fixing the doc 
> > >> rather than
> > >> the code.
> > >
> > > I understand this to mean that we should fix the doc to 
> > > reflect that
> > > `debug-on-error` is no longer explicitly non-nil when 
> > > --debug-init is
> > > passed.
> >
> > Yes.
> >
> > > But for the other issue, I think we should still make it so 
> > > that
> > > condition-case-unless-debug works with --debug-init as 
> > > advertised (and
> > > as it did in Emacs 29).  Do you agree?
> >
> > No, for the same reason: `condition-case-unless-debug` depends 
> > on the
> > value of `debug-on-error`, not on whether the user has asked 
> > some variant of
> > "show the debugger when there's an error".
>
> condition-case-unless-debug's docstring says "it does not 
> prevent
> debugging".  It also alludes to the value of debug-on-error, but 
> this
> doc could change as well :)

Yes, but it would bring us back to the problem of how to 
implementing it
without re-introducing the problem.

> I would argue, the intention of the programmer who used
> condition-case-unless-debug likely doesn't depend on the exact 
> value
> of debug-on-error, but rather, whether or not the user "wants" 
> to
> invoke the debugger when an error is signaled inside the body.
>
> > Another way to look at it:
> >
> > `--debug-init` is there so you can debug an error that breaks 
> > your init.
> > In the case of `condition-case-unless-debug`, the error 
> > presumably
> > doesn't break your init, since it is properly caught and 
> > handled.
>
> In my experience, condition-case-unless-debug seems to be often 
> used
> to "ignore" an error but not properly catch and handle it.  For
> example in the macroexpansion of `with-demoted-errors`, or in 
> the
> original place where I found this: Spacemacs, where it is used 
> to log
> an error message somewhere and proceed with installing the rest 
> of the
> packages that are needed even if one installation fails.
>
> For me, the major concern is, code which was intended to be 
> called
> during init and used condition-case-unless-debug to make its 
> errors
> non-fatal but debuggable, has broken in a backwards-incompatible 
> way,
> and also doesn't seem to have an obvious way forwards.

There's a simple way forward: set `debug-on-error`.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.