Codecov Report

Merging #5271 into master will increase coverage by 0.51%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #5271      +/-   ##
==========================================
+ Coverage   93.44%   93.95%   +0.51%     
==========================================
  Files         115      115              
  Lines       26163    26165       +2     
  Branches     2578     2578              
==========================================
+ Hits        24447    24583     +136     
+ Misses       1395     1259     -136     
- Partials      321      323       +2

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 43617a8...e253029. Read the comment docs.

@asottile
I think this should have waited on #5256 (comment).

@blueyed I think it is fine, we can always do a follow-up. 👍

And after giving this some more thought, we needed this fix because our test suite used 😄 emoji in the messages (my suggestion), so it might not even happen in the wild as I doubt people are putting emojis in exception messages.

Either way I think a follow up later (including a CHANGELOG adjustment) is fine. 👍

Sure. But there was also no reason to get this intermediate fix in fast - since we're not testing against Jython.

Anyway, it was just a remark - nothing too bad.

The current change is already ok for me (Jython works)...

As a note, what I came up with in the end is below (where obj_repr is a unicode string and left_count the index to be sliced):

try:
    if left_count > 0 and unichr(0xD800) <= obj_repr[left_count - 1] <= unichr(0xDBFF):
        left_count -= 1
except ValueError:
    # On Jython unichr(0xD800) will throw an error:
    # ValueError: unichr() arg is a lone surrogate in range (0xD800, 0xDFFF) (Jython UTF-16 encoding)
    # Just ignore it in this case.
    pass

I'm also a bit wary of pytest doing msg.rstrip() with a unicode identifier where msg is bytes on Python 2 (that could fail depending on the encoding used to automatically make the conversion from bytes to unicode) -- in: d94b4b0#diff-606d01f030119020f3e820547882b6b0R1003 (but as @nicoddemus mentioned, it's quite rare that exceptions have unicode at all, so, this may not happen all that much in practice).

I'm also a bit wary of pytest doing msg.rstrip() with a unicode identifier where msg is bytes on Python 2

You are right @fabioz, that could lead to problems.

I propose we wait a bit though, I'm not sure this will really be a problem in the wild. If it does we can always fix it with an actual problem as guide. 👍

Thanks everyone!