Only run single-character case conversion tests when unicode versions match #684
Conversation
Lysxia
force-pushed
the
test-case-mapping
branch
from
3b307df to
1154748
Compare
|
Nice! |
|
|
||
| #if MIN_VERSION_base(4, 15, 0) | ||
| -- Requires matching version of Unicode in base and text | ||
| testGroup "char case conversion" $ if T.unicodeVersion == G.unicodeVersion then [ |
There was a problem hiding this comment.
FWIW, it's very easy to get into state where those tests are not run at all. Any upgrade's to Unicode version used by text will cause these tests to not be run, as GHC is (usually) behind.
And you kind of want these tests to be run when Unicode stuff is changed...
There was a problem hiding this comment.
Yeah, that's a downside. But on the other hand having sporadical CI failures and maintaining t_{toUpper,toTitle}_exceptions is not confidence inducing either.
There was a problem hiding this comment.
That downside is why I took this long to decide to do this eventually. On the other hand, these tests will still run when CI is updated with the right GHC version. And when they do fail that will most probably mean that we need to update the list of exceptions in the test suite rather than an actual bug in the library.
Close #501
The tests are breaking again (though with low probability) because of Unicode version mismatches between text and base. This fixes that flakiness for good by recording the unicode version and comparing it with
GHC.Unicode.