codecov-commenter
commented
1 year ago Codecov Report
Merging #1923 (9312d87) into develop (d43f586) will decrease coverage by
0.12%. The diff coverage is50.00%.
@@ Coverage Diff @@
## develop #1923 +/- ##
===========================================
- Coverage 77.02% 76.90% -0.12%
===========================================
Files 58 58
Lines 6773 6795 +22
===========================================
+ Hits 5217 5226 +9
- Misses 1556 1569 +13 | Impacted Files | Coverage Δ | |
|---|---|---|
| web/main/models.py | 73.97% <25.00%> (-0.04%) |
:arrow_down: |
| web/main/export.py | 91.26% <66.66%> (-1.30%) |
:arrow_down: |
| web/main/admin.py | 70.87% <0.00%> (-1.19%) |
:arrow_down: |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
There's a subtle bug somewhere in the code that reserializes annotations in cases where it's possible to add a link annotation inside another tag and end up with the serializer trying to output:
As a consequence, any code that called this (either the docx export or the reading length calculation) would stop at this point due to an uncaught exception.
The problem is something introduced by our code, not due to bad markup in the original. But in practice I think it doesn't matter if we output
<a href=""><span>I'm unclosed</a></span>because both real browsers and the docx export code can deal with this fine.This PR also more strongly guards against uncaught exceptions in the serialization code in the context of calculating reading length. This calculation is not mission-critical and so there's no particular reason we need to be notified nor that the user should be prevented from continuing. Unknown errors will still otherwise propagate during exports, because we want to know about those. Lastly, the correct behavior when skipping calculations due to errors is to report 0 here, because a value of
Nonemeans it will try to recalculate every time, even if the content is unchanged.