Any downside to condition-case-unless-debug?

Any downside to condition-case-unless-debug?

[Eric Abrahamsen]

Hi,

I feel like a lot of my debugging recently has involved various ways of
working around `condition-case' and its swallowing of useful error
information. Is there any reason not to prefer
`condition-case-unless-debug' in new code?

Thanks,
Eric

Re: Any downside to condition-case-unless-debug?

[Glenn Morris]

Eric Abrahamsen wrote:

> Is there any reason not to prefer `condition-case-unless-debug' in new
> code?

It breaks ert. https://debbugs.gnu.org/11218

Re: Any downside to condition-case-unless-debug?

[Eric Abrahamsen]

Glenn Morris <rgm@gnu.org> writes:

> Eric Abrahamsen wrote:
>
>> Is there any reason not to prefer `condition-case-unless-debug' in new
>> code?
>
> It breaks ert. https://debbugs.gnu.org/11218

Oh, good to know, thanks. I guess I'll watch that bug report.

Re: Any downside to condition-case-unless-debug?

[Stefan Monnier]

> I feel like a lot of my debugging recently has involved various ways of
> working around `condition-case' and its swallowing of useful error
> information. Is there any reason not to prefer
> `condition-case-unless-debug' in new code?

IME, `condition-case-unless-debug' is used/useful mostly for
`condition-case` that catch "any error", most of which might be better
served by something like `with-demoted-errors`.

For those `condition-case` which catch one particular kind of error to
handle a particular expected situation (e.g. when catching `scan-error`
around a call to `forward-sexp`), `condition-case-unless-debug` rarely
makes sense.


        Stefan

Re: Any downside to condition-case-unless-debug?

[Eric Abrahamsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> I feel like a lot of my debugging recently has involved various ways of
>> working around `condition-case' and its swallowing of useful error
>> information. Is there any reason not to prefer
>> `condition-case-unless-debug' in new code?
>
> IME, `condition-case-unless-debug' is used/useful mostly for
> `condition-case` that catch "any error", most of which might be better
> served by something like `with-demoted-errors`.
>
> For those `condition-case` which catch one particular kind of error to
> handle a particular expected situation (e.g. when catching `scan-error`
> around a call to `forward-sexp`), `condition-case-unless-debug` rarely
> makes sense.

Yes, exactly! The cases I'm looking at are all cases that catch "any
error", often for no good reason (IMO). I'll try to tend towards patches
that remove `condition-case' altogether, rather than
`condition-case-unless-debug'.

Re: Any downside to condition-case-unless-debug?

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > I feel like a lot of my debugging recently has involved various ways of
  > working around `condition-case' and its swallowing of useful error
  > information. Is there any reason not to prefer
  > `condition-case-unless-debug' in new code?

If we wanted to have that behavior by default, then rather than change
lots of code gradually, how about making a new global flag
`debug-on-caught-error' which, when t, would make errors invoke
the debugger when they are about to be caught?

If `debug-on-caught-error' is `debug-on-error', then the value of
`debug-on-error' would control the behavior of errors about to be
caught.

-- 
Dr Richard Stallman (https://stallman.org)
Chief GNUisance of the GNU Project (https://gnu.org)
Founder, Free Software Foundation (https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)

Re: Any downside to condition-case-unless-debug?

[Eric Abrahamsen]

Richard Stallman <rms@gnu.org> writes:

> [[[ To any NSA and FBI agents reading my email: please consider    ]]]
> [[[ whether defending the US Constitution against all enemies,     ]]]
> [[[ foreign or domestic, requires you to follow Snowden's example. ]]]
>
>   > I feel like a lot of my debugging recently has involved various ways of
>   > working around `condition-case' and its swallowing of useful error
>   > information. Is there any reason not to prefer
>   > `condition-case-unless-debug' in new code?
>
> If we wanted to have that behavior by default, then rather than change
> lots of code gradually, how about making a new global flag
> `debug-on-caught-error' which, when t, would make errors invoke
> the debugger when they are about to be caught?
>
> If `debug-on-caught-error' is `debug-on-error', then the value of
> `debug-on-error' would control the behavior of errors about to be
> caught.

Would this help work around the ERT interference that Glenn referenced?