Add response_length to request_failure event

[cyberw]

Like with request_success... This breaks compatibility with existing tests that use event handlers with positional arguments. Maybe it is worth a bumped minor version, to indicate that to users.

[codecov]

Codecov Report

Merging #1144 into master will increase coverage by 0.5%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master    #1144     +/-   ##
=========================================
+ Coverage   75.12%   75.62%   +0.5%     
=========================================
  Files          18       20      +2     
  Lines        1833     1879     +46     
  Branches      280      290     +10     
=========================================
+ Hits         1377     1421     +44     
- Misses        390      391      +1     
- Partials       66       67      +1

Impacted Files	Coverage Δ
locust/contrib/fasthttp.py	89.01% <ø> (-1.16%)	⬇️
locust/clients.py	91.48% <ø> (ø)	⬆️
locust/stats.py	82.33% <100%> (ø)	⬆️
locust/core.py	84.81% <0%> (-0.41%)	⬇️
locust/exception.py	100% <0%> (ø)	⬆️
locust/main.py	35.24% <0%> (ø)	⬆️
locust/wait_time.py	100% <0%> (ø)
locust/util/deprecation.py	85% <0%> (ø)
locust/runners.py	65.1% <0%> (+1.37%)	⬆️

---

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 c16dc71...7d5eeeb. Read the comment docs.

[heyman]

Nice! The response_length should also be passed along to the log_error() call in on_request_failure on this line:

locust/locust/stats.py

Line 616 in 795b5a1

global_stats.log_error(request_type, name, exception)

Maybe it is worth a bumped minor version, to indicate that to users.

Yeah, I think that the next version we release should be 0.13 (the new wait time API have now also been merged).

[heyman]

Hmm, we might also need something like #1088, but for response_length

[cyberw]

Hmm, we might also need something like #1088, but for response_length

Responses that have an undefined response_length? Maybe... but I think we're ok with counting them as zero for now. Or do you feel this relates to the PR? As we dont calculate average response length, it doesnt really hurt that undefined response length cannot be separated from zero response times.

[heyman]

As we dont calculate average response length

We do calculate the average response length:

We've been using 0 for failures up until now (since it wasn't available in the event params), so it wouldn't make it more wrong to keep doing it, though maybe it would be worth fixing it now when it is available?

[cyberw]

Oh, I didnt see it in the StatsEntry object, so I thought it wasnt used. Maybe I should start using the web UI more :)

But it looks like it is already fixed by my changes?

[heyman]

But it looks like it is already fixed by my changes?

Ah, nice!

Then I guess we can merge this, and perhaps create a separate issue for excluding failures with None as response_length, from the average response size.