Fix/improve comparison of byte strings

[blueyed]

Fixes #5260.

[codecov]

Codecov Report

Merging #5267 into features will increase coverage by 1.16%.
The diff coverage is 100%.

@@             Coverage Diff              @@
##           features    #5267      +/-   ##
============================================
+ Coverage     92.07%   93.24%   +1.16%     
============================================
  Files           115      115              
  Lines         26159    26356     +197     
  Branches       2578     2623      +45     
============================================
+ Hits          24086    24575     +489     
+ Misses         1746     1465     -281     
+ Partials        327      316      -11

Impacted Files	Coverage Δ
testing/test_assertion.py	97.69% <100%> (+0.02%)	⬆️
src/_pytest/assertion/util.py	98.16% <100%> (+0.01%)	⬆️
testing/python/raises.py	88.51% <0%> (-6.09%)	⬇️
testing/test_capture.py	99.33% <0%> (+0.08%)	⬆️
src/_pytest/fixtures.py	97.93% <0%> (+0.13%)	⬆️
src/_pytest/pytester.py	86.66% <0%> (+0.14%)	⬆️
src/_pytest/capture.py	95.28% <0%> (+0.18%)	⬆️
testing/python/collect.py	99.14% <0%> (+0.21%)	⬆️
testing/test_runner.py	60.3% <0%> (+0.65%)	⬆️
testing/python/metafunc.py	95.22% <0%> (+2.94%)	⬆️
... and 4 more

---

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 c0e53a6...58b3d2a. Read the comment docs.

[blueyed]

Azure failure is unrelated.
(cannot restart it myself?!)

[asottile]

hmmm I thought that we wouldn't see this line, but only see the - b'spam' + b'eggs' bit?

[blueyed]

No - it just skips the "contains more items" now.
I think it is fine / useful, although it could mention that those a bytes / ordinals.
If so, how should it sound like then?
Or should we skip any handling (i.e. as a sequence) for bytes?

[asottile]

hmmm I'd like to still see the reprdiff since I think that's useful

[nicoddemus]

How about using the slice operator instead?

>>> b'foo'[0]
102
>>> b'foo'[0:1]
b'f'

We will then be able to see "At index 0 diff: b's' != b'e'", which is readable and has no downsides as far as I can see.

[blueyed]

As a special case for bytes on py3 then, or for sequences in general?

[nicoddemus]

Yes, as a special case for bytes on py3. For sequences it is not clear to me it would be better than what we have now, at least for me.

[blueyed]

I do not feel strong enough about this myself currently, closing for now.
Let's keep it for later - almost there.. ;)

[nicoddemus]

Let's keep it for later - almost there.. ;)

By "later" you mean after dropping Python 2 in 4.6/4.7?

[nicoddemus]

Gentle ping :)

[blueyed]

Busy pong.. ;)
Feel free to pick it up or close it to get off your radar.. :)

[nicoddemus]

Superseded by #5495