figures
figures copied to clipboard
sindresorhus
Fallback symbols are not covered by tests in Unicode-capable environments, including CI
Observed at https://github.com/sindresorhus/figures/pull/93#discussion_r939671383 , the test for replaceSymbols verifies the absence of replacements (in a Unicode-capable environment) OR replacement with fallback symbols (in other environments), but not both. It would be good to update the tests for more complete coverage, or failing that to at least update CI to include less capable environments.
Yes, but I may need help with this one. Is there a standard approach in these packages to mocking dependencies such that isUnicodeSupported() can be forced to return false inside a test? And if not, then what would be your alternative recommendation?
Good point.
Not sure how @sindresorhus feels about this, but I personally try to avoid mocking whenever possible because:
- It can end up creating complex test setups
- It makes the tests less similar than the environment the users actually experience, which makes the tests less useful, and can miss some bugs
In this situation, I think what'd be best would be to run the tests in CI in an environment where isUnicodeSupported() returns false. In principle, Windows machines on GitHub actions use ConsoleHost (through cmd.exe) which uses its own charset and has issues displaying many Unicode characters. Unfortunately , isUnicodeSupported() returns true when the CI environment variable is set. What was the reason behind this @sindresorhus?
If changing is-unicode-supported is not an option, I am wondering whether setting CI to an empty string in the GitHub actions workflow file for the Windows environment would work? :thinking:
Unfortunately , isUnicodeSupported() returns true when the CI environment variable is set. What was the reason behind this @sindresorhus?
I don't remember. It's definitely not the correct behavior.
https://github.com/sindresorhus/is-unicode-supported/releases/tag/v2.0.0
