PowerShell linter does not count errors properly

[efrecon]

Describe the bug
The powershell linter does not count errors properly. Instead of reporting the number of errors, it report the number of erroneous files.

To Reproduce
Steps to reproduce the behavior:

1. Run the powershell linter on a file with several warnings/errors
2. Megalinter reports "1" error.

Expected behavior
The MegaLinter should count the number of errors and probably report warnings as non-blocking errors.

note
I believe this is because the descriptor does not provide a cli_lint_errors_count and cli_lint_errors_regex or similar.

[echoix]

Does the tool itself have a way of reporting it? It's a bit harder to be smarter than the tool ;)

[efrecon]

Not directly, not that I know. But, as I understand the implementation of the MegaLinter, the two descriptor keys that I mentioned above are able to count the number of occurrences of matches for the cli_lint_errors_regex regular expression. Is it so?

In that case, I think that it should be possible to control the order of the columns output from Format-Table (see #3318) and arrange for the severity to be placed first. We could then count the occurrences of ^[A-Za-z]*Error (or ^Warning for non-blocking errors).

[nvuillam]

Indeed cli_lint_errors_count and cli_lint_errors_regex are the way to count errors using output text :)

@efrecon please feel free to add them, even in your current PR about Powershell if you want :)

[efrecon]

Merci @nvuillam. I've started to look at that but stumbled upon the following:

If I understand the code correctly, when the total_errors is 0 as of this line -- because there are only warnings or information (possible levels here) -- the python code will issue a warning. And, since there was some output, this will end up being reported as 1 error (there was 1 file with some problems).

• Is my understanding correct?
• Is this correct behaviour?
• How can I support "non-breaking" errors, e.g. Warning in the severity table linked above.

[echoix]

I'm not sure I understand your message correctly.

If I'm not mistaken, the configuration key DISABLE_ERRORS_LINTERS enables you to have a linter run, but if the linter reports errors, instead of failing the job, it will only report it as warnings.

https://megalinter.io/latest/configuration/
https://megalinter.io/latest/configuration/#activation-and-deactivation

Is it a bit what you are trying to achieve?

Or you are trying to "upgrade" the existing linter to be able to report both kinds (warnings AND errors) when it does (or doesn't) currently report it when used by itself manually.

[efrecon]

I needed a helping hand. This is now implemented as part of #3318. The implementation places the severity at the beginning of the lines and count the known severity levels in the output. This means that even messages with a severity of Information will report errors, but it is possible to control the level of severity through the configuration file pointed at by POWERSHELL_POWERSHELL_CONFIG_FILE.

I have run the tests as here. They pass. Is there anything I need to do for this to be considered for approval and merge?

[nvuillam]

@efrecon if the CI jobs pass, I'll merge the PR :)

[efrecon]

@nvuillam They.. didn't. One line too long, the rest being internal: a network error and a "no space on device error"

[nvuillam]

Let's cross fingers for the next run :)

I updated your branch from main, the no space left on device has been solved in another PR

[nvuillam]

Appears to be solved by #3318 :)