SR-331: Diagnostic notes and fixits for unicode confusables #9070
Conversation
|
Fantastic! @swift-ci please smoke test |
|
Since the approach is roughly the same (and the PR comments from last time have been addressed), over to @lattner for if this is what he had in mind back then. |
|
@robinkunde Please lint the Python script and fix the errors. |
|
@CodaFi I can't seem to find a clean way to fix the error where it complains about the docstring line length. The leading and closing line for def file header are supposed to be 80 columns wide. The python linter insists lines be at most 79 wide. For now I've broken it up into multiple write statements, but it looks ugly. If someone knows a better way, please chime in. |
|
Sometimes the linter isn't always right - can we add specific exceptions for this @hughbe? |
|
@swift-ci please smoke test |
|
Just to check: @swift-ci please test source compatibility |
|
There's no reason to delay the inclusion of this feature any further. Thank you @robinkunde for carrying this over the goal line, and @amackworth for the original patch. Don't forget to resolve the SR. I'm going to close the original pull request as well. ⛵️ |
Addresses SR-331 by adding diagnostic notes and fixits for confusable unicode characters. The confusables check only runs after the lexer encounters an invalid character or an identifier fails to resolve.
With @amackworth permission, I took over the code he wrote for #732. From the discussion there, it seems like the PR stalled due to disagreements over whether to use gyb for compile time code gen. I reworked the code to address most of the feedback and removed gyb usage. The flow is now as follows:
This way, the script only has to be rerun if we broaden the list of covered confusables or Unicode releases a new version (happens about once a year).