Add fontawesome regression tests#33
Conversation
|
@jneem I'm not sure what to make of the fact that the bug I found in my other testing system doesn't reproduce here. My other system uses linesweeper However, when I reset to branch off of |
|
Interesting! I'm not sure what could cause this, but I'll see if I can reproduce on my end with the released crate. In general, I like the direction of this PR although I have reservations about serde_yaml since it's officially unmaintained... |
|
Ok. I didn't know serde_yaml was unmaintained. My default would have been serde_json, but it seems like yaml is more readable for this kind of declaration file. I think that leaves toml or json as the fallback options. Do you have any strong feelings about either of those? |
|
No strong preferences, really. It's a pity because small examples like this are nice in yaml, but the format as a whole is so complex that the rust support just isn't great |
|
Once I got linesweeper incorporated into my production system, the error for these cases also didn't reproduce. So I'm going to close this PR for now. The additional snapshot testing might be interesting to have, but there don't seem to be any interesting test cases for it at the moment. So maybe I'll circle back to something like this if/when another bug surfaces where this kind of regression testing would be beneficial. |
This is a work-in-progress. It might be scrapped, because it doesn't reproduce the bugs found elsewhere. Also, it may be good to cleanup and refactor before merging, if it's useful at all (which I'm now doubting).
This refactors the regression integration tests in order to add some from Font Awesome.
There are three tests of
binary_opwithDifferenceandNonZero. When I run them in my other test system, the images look like this:Base:

Subtract this (

Difference):Expect this:

Actually get this:
(Note that the images I've uploaded here are not the real ones from my testing system. Some of these are screenshots of the rendered SVGs.)
As evident by the snapshots added in this PR, linesweeper produced the expected results instead of the buggy results.
I haven't yet figured out what's different between my other testing system and my attempt to import these as regression test cases into the linesweeper repo.
However, I will note that I ran 5,552,768 test cases, and these are the only three that failed! Awesome.