Revert change to traceback repr

[Zac-HD]

This PR ammends #7274, undoing an incidental change to the output string and therefore fixes #7534.

cc @nicoddemus @hroncok

[The-Compiler]

Would be good to have a test to prevent this kind of regression in the future? It looks like testing/code/test_excinfo.py would be the best place.

[Zac-HD]

I did wonder about that, but couldn't think of a test that would catch similar issues without being rather fragile to non-problems like the names of parent directories. Personally I'm OK with relying on low code churn here, but if anyone wants to push a test to this branch be my guest 🙂

[nicoddemus]

I did wonder about that, but couldn't think of a test that would catch similar issues without being rather fragile to non-problems like the names of parent directories. Personally I'm OK with relying on low code churn here, but if anyone wants to push a test to this branch be my guest

I think a test is important here, we would just need to use a regex to make the test independent from the actual file path.

[nicoddemus]

I think we should show the non-repr of the path here:

Suggested change

	return " File %r:%d in %s\n %s\n" % (
	return " File %s:%d in %s\n %s\n" % (

[The-Compiler]

I don't think so - Python also quotes the path in its tracebacks:

Traceback (most recent call last):
  File "test.py", line 1, in <module>
    raise Exception

[The-Compiler]

Ah, though Python uses " in any case:

  File "test"foo.py", line 1, in <module>

so I suppose we could do File \"%s\" here instead (and drop the str(...) again?).

[nicoddemus]

I don't think so - Python also quotes the path in its tracebacks:

Ahh you are right. 👍

so I suppose we could do File "%s" here instead (and drop the str(...) again?).

Right on. 😁

[bluetech]

@The-Compiler

File "test.py", line 1, in

The code seems to (before my oversight) do

File "test.py":1 in

Which looks kinda weird. Should we indeed change it to match the Python format, i.e File "%s", line %d, in %s?

Though that would break werkzeug as well, so not for this PR.

[The-Compiler]

Oh, I hadn't noticed that. I guess let's just stay with the previous format for now then?

[nicoddemus]

Actually I noticed previously this used ' instead of ", based on the failure reported in #7534:

        traceback = "".join(str(line) for line in baz_exc.traceback)
>       assert "bb.py':1" in traceback  # a bit different than typical python tb

(Nevermind this, replied to a stale comment hehehe)

[Zac-HD]

@nicoddemus - I've pushed a final small tweak, because "'%s'" % (any_string,) is not quite the same as "%r" % (any_string,) - e.g. when any_string contains a ' single-quote.

This last patch therefore restores the previous behaviour exactly, and adds a comment explaining why we're not matching Python. IMO if we're going to make a change at all, we should do it to exactly match Cpython and be done with it; but that's a topic for later.

[The-Compiler]

Looks great! Not sure if it needs a changelog entry or not, though?

[nicoddemus]

Looks great! Not sure if it needs a changelog entry or not, though?

I think it makes sense because the change was released in 6.0.0rc1, and I gather we will keep the release candidate changelogs in the final version.

[nicoddemus]

This last patch therefore restores the previous behaviour exactly, and adds a comment explaining why we're not matching Python.

Awesome. 👍

[The-Compiler]

One step closer to a final 6.0 then! 🎉 Thanks everyone!