Backtrace mode

Backtrace mode

[Gemini Lasswell]

Hi All,

I've finished everything I set out to add to Backtrace mode, and have
pushed it to scratch/backtrace-mode. Documentation, NEWS and tests are
all done, so please have a look at it and let me know if you find
anything that should be fixed before I merge it.

Notable changes:

- It's now possible to toggle both local variable display and printing
with 'print-circle' for the entire buffer instead of just the current
frame, using the C-u prefix with 'v' and '#' respectively.

- C functions in backtraces now get links to their source code. Links
are also created for the reading position of 'eval-buffer' and
'eval-region'.

- I've added a new function 'cl-print-to-string-with-limit' to
cl-print.el, which tries to get a printed representation to fit in a
specified number of characters using 'print-length' and 'print-level'.
Backtrace mode is using it and I plan to use it for results printing
in Edebug, the Lisp Debugger and ERT, hopefully putting an end to
performance problems caused by excessively long lines in those modes.

- You can now expand all the '...'s in a backtrace frame at once using
backtrace-expand-ellipses, bound to '.' and with the C-u prefix expand
them without putting a limit on the expansion.

- In Edebug backtraces, three new commands are available.
backtrace-goto-source, bound to 's', works on frames in
Edebug-instrumented code, which are marked with a '>' on the left side.
I've also added edebug-backtrace-show-instrumentation and
edebug-backtrace-hide-instrumentation, to show and hide Edebug's
functions in the backtrace.

- The function 'backtrace' now uses Backtrace mode's printing and is now
located in backtrace.el instead of subr.el. The functions
'backtrace-frame' and 'backtrace-frames' remain in subr.el so that
Backtrace mode doesn't have to be loaded for programmatic access to the
backtrace from Lisp.

Re: Backtrace mode

[John Wiegley]

>>>>> "GL" == Gemini Lasswell <gazally@runbox.com> writes:

GL> I've finished everything I set out to add to Backtrace mode, and have
GL> pushed it to scratch/backtrace-mode. Documentation, NEWS and tests are all
GL> done, so please have a look at it and let me know if you find anything
GL> that should be fixed before I merge it.

Sounds useful!

-- 
John Wiegley                  GPG fingerprint = 4710 CF98 AF9B 327B B80F
http://newartisans.com                          60E1 46C4 BD1A 7AC1 4BA2

Re: Backtrace mode

[Eli Zaretskii]

> From: Gemini Lasswell <gazally@runbox.com>
> Date: Fri, 27 Jul 2018 13:13:30 -0700
> 
> I've finished everything I set out to add to Backtrace mode, and have
> pushed it to scratch/backtrace-mode. Documentation, NEWS and tests are
> all done, so please have a look at it and let me know if you find
> anything that should be fixed before I merge it.

Thanks.  A few minor nits regarding documentation:

 . The first line of each NEWS entry should ideally be a single
   sentence; all the following sentences should begin on the second
   line.
 . Leave 2 spaces between sentences.
 . The closing brace of a cross-reference should be followed by a
   period or a comma.
 . The last (5th) argument of a cross-reference should be the official
   title of the manual to which you point, as stated in the title
   page.
 . When making cross-references to the current file, there's no need
   to specify that file's name by using the 4-argument cross-reference
   form.

Re: Backtrace mode

[Gemini Lasswell]

Eli Zaretskii <eliz@gnu.org> writes:

> Thanks.  A few minor nits regarding documentation:

Thanks for the feedback.  I've taken care of the nits,
and I'll merge the branch in a day or two.

Re: Backtrace mode

[Clément Pit-Claudel]

Hi Gemini,

I just gave the new code a try.  This is really neat!  Thanks a lot :)
Did you consider using V instead of C-u v to toggle all locals?

Clément.

On 2018-07-27 16:13, Gemini Lasswell wrote:
> Hi All,
> 
> I've finished everything I set out to add to Backtrace mode, and have
> pushed it to scratch/backtrace-mode. Documentation, NEWS and tests are
> all done, so please have a look at it and let me know if you find
> anything that should be fixed before I merge it.
> 
> Notable changes:
> 
> - It's now possible to toggle both local variable display and printing
> with 'print-circle' for the entire buffer instead of just the current
> frame, using the C-u prefix with 'v' and '#' respectively.
> 
> - C functions in backtraces now get links to their source code. Links
> are also created for the reading position of 'eval-buffer' and
> 'eval-region'.
> 
> - I've added a new function 'cl-print-to-string-with-limit' to
> cl-print.el, which tries to get a printed representation to fit in a
> specified number of characters using 'print-length' and 'print-level'.
> Backtrace mode is using it and I plan to use it for results printing
> in Edebug, the Lisp Debugger and ERT, hopefully putting an end to
> performance problems caused by excessively long lines in those modes.
> 
> - You can now expand all the '...'s in a backtrace frame at once using
> backtrace-expand-ellipses, bound to '.' and with the C-u prefix expand
> them without putting a limit on the expansion.
> 
> - In Edebug backtraces, three new commands are available.
> backtrace-goto-source, bound to 's', works on frames in
> Edebug-instrumented code, which are marked with a '>' on the left side.
> I've also added edebug-backtrace-show-instrumentation and
> edebug-backtrace-hide-instrumentation, to show and hide Edebug's
> functions in the backtrace.
> 
> - The function 'backtrace' now uses Backtrace mode's printing and is now
> located in backtrace.el instead of subr.el. The functions
> 'backtrace-frame' and 'backtrace-frames' remain in subr.el so that
> Backtrace mode doesn't have to be loaded for programmatic access to the
> backtrace from Lisp.
> 
> 

Re: Backtrace mode

[Clément Pit-Claudel]

On 2018-07-27 16:13, Gemini Lasswell wrote:
> - C functions in backtraces now get links to their source code. Links
> are also created for the reading position of 'eval-buffer' and
> 'eval-region'.

A small potential bug: if I install ido-completin-read+ emacs -Q, then running M-: (debug) produces a backtrace in which call-interactively isn't highlighted:

- Without ido-cr+ (all lines have a hyperlink to their sources):

  Debugger entered: nil
    eval((debug) nil)
    eval-expression((debug) nil nil 127)
    funcall-interactively(eval-expression (debug) nil nil 127)
    call-interactively(eval-expression nil nil)
    command-execute(eval-expression)
  
- With ido-cr+ (lines marked with ‘!’ don't have hyperlinks):

  Debugger entered: nil
    eval((debug) nil)
    eval-expression((debug) nil nil 127)
    funcall-interactively(eval-expression (debug) nil nil 127)
!   #<subr call-interactively>(eval-expression nil nil)
    apply(#<subr call-interactively> eval-expression (nil nil))
    call-interactively@ido-cr+-record-current-command(#<subr call-interactively> eval-expression nil nil)
    apply(call-interactively@ido-cr+-record-current-command #<subr call-interactively> (eval-expression nil nil))
!   call-interactively(eval-expression nil nil)
    command-execute(eval-expression)

I imagine this is due to advice? 

    :around advice: ‘call-interactively@ido-cr+-record-current-command’

Cheers,
Clément.

Re: Backtrace mode

[Gemini Lasswell]

Clément Pit-Claudel <cpitclaudel@gmail.com> writes:

> I just gave the new code a try.  This is really neat!  Thanks a lot :)
> Did you consider using V instead of C-u v to toggle all locals?

Hi Clément,

Glad you like it! No, it didn't occur to me to use V. But there's also
backtrace-toggle-print-circle which is currently also using C-u to make
it work on the whole buffer, and which would need a different binding if
the idea is for the shift key to mean do the whole buffer.

I bound backtrace-toggle-print-circle to # because p is used by
backtrace-backward-frame and c by debugger-continue. The number of #'s
in the printed representations of circular objects changes with the
value of print-circle, so I thought it would be mneumonic.

Gemini

Re: Backtrace mode

[Gemini Lasswell]

Clément Pit-Claudel <cpitclaudel@gmail.com> writes:

> A small potential bug: if I install ido-completin-read+ emacs -Q, then running M-: (debug) produces a backtrace in which call-interactively isn't highlighted:
>
> I imagine this is due to advice? 

Thanks for the bug report, and the correct diagnosis. I'll see what I can do about it.

Re: Backtrace mode

[Clément Pit-Claudel]

On 2018-08-03 13:14, Gemini Lasswell wrote:
> Clément Pit-Claudel <cpitclaudel@gmail.com> writes:
> 
>> A small potential bug: if I install ido-completin-read+ emacs -Q, then running M-: (debug) produces a backtrace in which call-interactively isn't highlighted:
>>
>> I imagine this is due to advice? 
> 
> Thanks for the bug report, and the correct diagnosis. I'll see what I can do about it.

Thanks for the cool improvements :)