Random seg fault while checking medium sized repo through docker

[marty30]

Run command: docker run --rm --volume=/var/lib/jenkins/workspace/path/to/repository:/check mstruebing/editorconfig-checker ec

.ecrc file (slightly redacted)

{
  "Version": "2.4.0",
  "Verbose": true,
  "Debug": false,
  "IgnoreDefaults": false,
  "Exclude": [
    ".*/target/.*",
    "target/.*",
    "schema.json$",
    "\\.idea/.*",
    "\\.editorconfig$",
    "dependencies_.*\\.txt",
    "docker/.*"
  ],
  "Disable": {
      "EndOfLine": false,
      "Indentation": false,
      "IndentSize": false,
      "InsertFinalNewline": false,
      "TrimTrailingWhitespace": false,
      "MaxLineLength": true
    }
}

Sometimes it fails on our Jenkins node running docker. The host is a RHEL7 machine.

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x28 pc=0x5a33b2]
goroutine 1 [running]:
github.com/editorconfig-checker/editorconfig-checker/pkg/files.GetFiles.func2(0xc000323b70, 0xd, 0x0, 0x0, 0x63baa0, 0xc0000a19e0, 0x0, 0x0)
	/ec/pkg/files/files.go:103 +0x52
path/filepath.walk(0xc0000b6c60, 0x6, 0x63e348, 0xc00011e4e0, 0xc0004cfcc0, 0x0, 0xc0000c6420)
	/usr/local/go/src/path/filepath/path.go:434 +0x226
path/filepath.Walk(0xc0000b6c60, 0x6, 0xc000115cc0, 0x0, 0x63bce0)
	/usr/local/go/src/path/filepath/path.go:501 +0x113
github.com/editorconfig-checker/editorconfig-checker/pkg/files.GetFiles(0x0, 0x5fb4bf, 0x5, 0xc0000b6a90, 0x5, 0x1, 0xc0000a3d40, 0xd, 0xd, 0x7438e0, ...)
	/ec/pkg/files/files.go:102 +0x5e5
main.main()
	/ec/cmd/editorconfig-checker/main.go:107 +0x318

Any suggestions?

[mstruebing]

What do you mean by sometimes? What's the difference when it fails and when not?

[marty30]

I have not yet noticed any difference. Still trying to figure that one out. I believe that re-running the same pipeline (so the same code) can pass the first time and fail the next, or the other way around.

[mstruebing]

Would be great if you could confirm this, and if this is the case.
Additionally, if that is the case it would be great if you could provide some steps to reproduce it, even if it only happens occasionally.

[marty30]

I have done some more research. Could it have to do with one process writing to a file while the editorconfig-checker is running? The panic seems to occur on Don't add /check/target/checkstyle-cachefile to be checked and we run checkstyle with maven at the same time.

[mstruebing]

Could be some kind of race condition if editorconfig-checker tries to read the file while it's not present anymore I think.

[marty30]

I now have proof that it without making any changes to the code, the exact same command fails first, and passes after. For now I added a retry mechanism to solve the issue for me. There are multiple processes that are running at the same time however (building the maven project, running checkstyle and running editorconfig-checker). So a race condition might very well be possible. Please let me know if you are able to resolve this issue or what specific information I can provide to help out. I have no Go experience, so it is hard to do any static analysis.

[mstruebing]

It would be very great to make this reproducible somehow locally.
If anyone is willing to provide a PR which can simulate that failure I can have a look.