[bug #55109] [PATCH] s.tmac: Add the name of the calling macro to a diagnostic

[G. Branden Robinson]

Update of bug #55109 (project groff):

                Severity:              3 - Normal => 1 - Wish               
                  Status:                    None => Wont Fix               
             Assigned to:                    None => gbranden               
             Open/Closed:                    Open => Closed                 

    _______________________________________________________

Follow-up Comment #1:

I think this patch has its heart in the right place but doesn't do quite the
right thing.

See bug #59290 for a more philosophical discussion of what I think we want
instead.

I'm attaching my proposed alternative, which is lower-impact in terms of line
count and the spread of impact through the file.

Here's an example of the output it produces.


$ printf '.TL\n' | ./build/test-groff -Tutf8 -z -ms
s.tmac: error: diversion open while ejecting page (last file seen: <standard
input>) (recovering)
$ printf '.TL\n' > /tmp/empty-title.ms && ./build/test-groff -Tutf8 -z -ms
/tmp/empty-title.ms
s.tmac: error: diversion open while ejecting page (last file seen:
/tmp/empty-title.ms) (recovering)
$ printf '.DS\n.DS\n.DE\n' > /tmp/unbalanced-diversion.ms &&
./build/test-groff -Tutf8 -z -ms /tmp/unbalanced-diversion.ms
s.tmac: error: diversion open while ejecting page (last file seen:
/tmp/unbalanced-diversion.ms) (recovering)
$ cat /tmp/two-TL.ms 
.TL
This Is My Title
.TL
There Are Many Like It
.PP
But this one is mine.
$ ./build/test-groff -Tutf8 -z -ms /tmp/two-TL.ms
s.tmac:/tmp/two-TL.ms:3: error: TL is not allowed more than once


We can see particularly from the final example above that the name of the
macro emitting the diagnostic is not what is of crucial importance; a
diagnostic message that accurately identifies the problem, accompanied by the
file name and location within, is what will help the user to remedy the
problem.

As it happens, the GNU Coding Standards has excellent advice in this domain. 
https://www.gnu.org/prep/standards/html_node/Errors.html

Linus Torvalds once said that the foregoing document should be set on fire and
the ashes urinated upon, a comment oft repeated by his many fans.  I am sure
the quality of the diagnostic messages those fans write speaks for itself.  Or
would, if such messages existed at all.  But in their ethos, Ken Thompson's
infamous '?' (from AT&T "ed") is one character of user communication too
many.

Per code review on the submitted patch, I regard the many scattered identical
changes of "@error:" to "@error: \\$0" and similar to be a code smell.  If
that's really what you wanted, the better thing to do would, I suspect, be to
interpolate @error as a string, which would preserve $0 and then @error could
interpolate that itself.  See the "Strings" node in the groff Texinfo manual. 
"Interpolating a string does not hide existing macro arguments."

Closing as "won't fix", but not because the patch wasn't well-intentioned.

(file #50005)
    _______________________________________________________

Additional Item Attachment:

File name: s.tmac.55109.diff              Size:1 KB
    <https://file.savannah.gnu.org/file/s.tmac.55109.diff?file_id=50005>



    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?55109>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/