Wrong underlining of words starting with markup

[ed359]

Describe the bug
When spell checking a word with an accented character at the beginning, typed using a latex command, the initial accented letter is omitted in the rule reporting a misspelled word.

Steps to reproduce
Type the word \v{s}ekki (which is šekki) into a latex document, with spell checking for English enabled and observe that "ekki" gets underlined, but not the entire word.

Expected behavior
Since šekki is not an English word it should be underlined, including the first character

Sample document
If the bug occurs for a specific document (e.g. LaTeX), please paste it here. If your document is very long or confidential, please create and attach a smaller example for which the bug still occurs so that we can reproduce it.

\documentclass{minimal}
\usepackage[T1]{fontenc}

\begin{document}
Šekki Štefancovi\v{c}.
Štefancovi\v{c} šekki.

\v{S}ekki \v{S}tefancovi\v{c}.
\v{S}tefancovi\v{c} \v{s}ekki.
\end{document}

LTeX configuration
Please paste all configuration settings starting with ltex. from your settings.json. You can help us by temporarily removing some irrelevant settings from your settings.json and see if the bug still occurs.
N/A

LTeX LS log
Paste the log (output) of LTeX LS here:

FINE: Checking the following text in language 'en-US' via LanguageTool: " \n\n\u0160ekki \u0160tefancovi\u010D. \u0160tefancovi\u010D \u0161ekki.\n\n\u0160ekki \u0160tefancovi\u010D. \u0160tefancovi\u010D \u0161ekki. "

Version information
List here the version information of the relevant software.

• Operating system: Windows 10
• ltex-ls: 12.1.0-alpha1 (built from source)
• Java: openJDK 11 LTS

Additional context/information

The log appears to show the correct strings are being sent to the server: \u0160 is Š and \u0161 is š. So something is probably going wrong outside of the (new) latex accent command parsing.

[valentjn]

There is a bug in the computation of the length of the underlining (which also determines the word X in the message 'X': Possible spelling mistake found.). The checking itself is not affected (i.e., if you use a word LanguageTool knows, there won't be a diagnostic). The Use '...' quick fixes display the correct suggestions; however, the wrong word will be replaced when applying them. The Add 'X' to dictionary quick fix will add the wrong word (without the first letter) to the dictionary, but as the dictionary is compared against the underlined words, this will still result in the diagnostic to be resolved. The other quick fixes are working as intended.

[ed359]

Ok, it's good to know that the underlying spell checking is correct, but the user interface reporting 'X': Possible spelling mistake found. is where the error is. Thank you.

[valentjn]

Thanks for the report, this should be fixed.

[ed359]

The new tests in 6317adc do not pass for me:

[ERROR] Failures: 
[ERROR]   LatexAnnotatedTextBuilderTest.testMathMode:347 30 not smaller than 30 ==> expected: <true> but was: <false>
[ERROR]   LatexAnnotatedTextBuilderTest.testTextMode:207 expected: <0> but was: <5>
[ERROR]   DocumentCheckerTest.testLatex:182 expected: <?ekki> but was: <ekki>

[valentjn]

They pass in GitHub Actions (and for me), so it's definitely working.

Best guess is that you forgot to update your local installation of ltexls-languagetool-patch (see second task in .vscode/tasks.json).

[ed359]

I wasn't using the tasks in tasks.json, so indeed that was the problem. Thank you for helping me run the tests. I'm on Windows and the build task fails with an error similar to the one described here: https://stackoverflow.com/questions/6347985/cannot-run-maven-using-mvn-d-argument-within-microsoft-powershell-but-works.

The issue is that powershell needs some quoting in the command string, namely '-Dmaven.test.skip=true' instead of -Dmaven.test.skip=true. If this change doesn't affect other OSes then it's probably a good idea in tasks.json.

[valentjn]

Fix released in 12.1.0.