False positive on CSpell due to @cspell/dict-win32 not being recognized

[mwarres]

Describe the bug
When we run CSpell through MegaLinter in CI, "nanoserver" and "ltsc", which are words in the CSpell's win32 dict, come up as false positives despite us having installed and configured the win32 dict.

To Reproduce
Steps to reproduce the behavior:

1. Run yarn init.
2. Run yarn add --dev @cspell/dict-win32@2.0.1.
3. Paste in a minimal CSpell config that includes the CSpell win32 dict.
4. Paste in a minimal MegaLinter config.
5. Make a file containing the strings "ltsc" and "nanoserver".
6. Make a GitHub Action test workflow to run MegaLinter.
7. Push to GitHub and note that CSpell incorrectly fails on the words "ltsc" and "nanoserver".

Here is a minimum reproduction

Here is the PR where we had this problem: ScribeMD/docker-cache#475

Expected behavior
Expect CSpell to pass on the words "ltsc" and "nanoserver".

Thank you for your help!!!

[Kurt-von-Laven]

Notably, the issue does not occur when running CSpell directly (e.g., npx cspell@6.31.1 .). As a workaround, one can install the unbundled dictionary in question using a pre-command in one's MegaLinter config:

SPELL_CSPELL_PRE_COMMANDS:
 - command: "npm install @cspell/dict-win32@2.0.1"

Does anyone have any theories as to why CSpell wouldn't be able to import a dictionary when run through MegaLinter?

[nvuillam]

I see you found a workaround, I close the issue :)
Please reopen if you think something can be done within MegaLinter (apart from adding all dictionaries by default in the docker image)

[Kurt-von-Laven]

Duplicate #1572 since this is also caused by the fact that CSpell can't see Yarn dependencies when not run via Yarn. This issue would also be addressed by accepting arrays in _CLI_EXECUTABLE, thereby allowing users to run CSpell via Yarn. Still open to better ideas.