Skip directories can't be scanned instead of aborting the whole check.

[spacewander]

Assumed we have directory structure like this: . ├── A │   └── B │   └── C └── D

If directory B could not be opened, previously luacheck will abort. Thus we have to run luacheck separately for directory C and D, and more.

With this patch, luacheck will continue to scan directory C and D.

[mpeterv]

It's a good idea to have this directory skipping behaviour, but simply writing the error to stderr doesn't work very well, other errors are more nicely formatted, and formatting can be different based on options (e.g. it could be XML when using JUnit formatter).

So extract_files should return all the errors somehow, and the filenames as well, and the calling function should handle them. I would keep returning an array of paths as the first return value, whether they had an error or not, and then also return a sparse array mapping indexes from the first table to error messages if for that path directory expansion failed. But I'd merge another implementation with the same behaviour, too.

[spacewander]

@mpeterv I will change this pull request once I have enough free time.

[codecov-io]

Codecov Report

Merging #159 into master will decrease coverage by 0.02%. The diff coverage is 61.53%.

@@            Coverage Diff             @@
##           master     #159      +/-   ##
==========================================
- Coverage   97.77%   97.75%   -0.03%     
==========================================
  Files          35       35              
  Lines        4819     4820       +1     
==========================================
  Hits         4712     4712              
- Misses        107      108       +1

Impacted Files	Coverage Δ
src/luacheck/fs.lua	90.42% <50%> (-1.07%)	:arrow_down:
src/luacheck/runner.lua	87.73% <80%> (+0.05%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 43d4645...54226be. Read the comment docs.

[spacewander]

@mpeterv I have just squashed my commits with the latest change and rebased it to include the Windows tests fix.

[mpeterv]

@spacewander I think there is a minor issue, I've commented on the diff. I'd merge and fix it myself but a bit busy now unfortunately.

[spacewander]

@mpeterv Good catch! I just submitted a new commit to fix the handling of initial scan call.

[mpeterv]

LGTM, thanks!