Fix git dirty check for revision_mode=scm

[aagor]

• Check whole git repository for dirtyness. For the commit hash retrieved, the whole repository is already taken into consideration. This solves the inconsistency between the commit hash retrieval and dirty check.
• Improve error message.

Changelog: BugFix: Check whole git repository for dirtyness when using revision_mode=scm
Docs: omit

• Refer to the issue that supports this Pull Request.
• If the issue has missing info, explain the purpose/use case/pain/need that covers this Pull Request.
• I've read the Contributing guide.
• I've followed the PEP8 style guides for Python code.
• I've opened another PR in the Conan docs repo to the develop branch, documenting this one.

[CLAassistant]

All committers have signed the CLA.

[memsharded]

I have some concerns that this might be breaking for users that have the repository dirty with some temp files outside of the conanfile subfolder, and suddenly this stops working. Having those temporary files in the root repo wouldn't be a big problem in most cases, isn't it?

Is this check really that important to protect at the repository level?

[aagor]

No, it isn't, at least not for me and not now.
But what is the use-case of using the whole git repo for the ref (as opposed to scm_folder), and having dirty stuff lying around then?

I saw this inconsistency during investigation of different methods for revision_mode and git source retrieval.
If it shouldn't be changed, we should at least add a comment to the source code, that it's the way it currently is on purpose and a link to that PR.

[aagor]

I think I have another issue that affects revision_mode for both scm_folder and scm (only if the change above gets not merged).

The dirty check should ensure that the git repository is clean below the folder defined by self.folders.root.
Otherwise you can miss dirty files that are used by Conan, e.g. in this example. E.g. changes in common folder pass the dirty check even if they are clearly relevant.

For revision_mode= "scm_folder", this is also relevant for the commit that gets written into conandata.yml. When something changes only at the common folder, you record the old commit from the hello or bye subfolder, which leads to the fact conan build does checkout something old.

Should I open a separate issue for that?

[aagor]

Hi @memsharded,

do you have any feedback on my questions above?
Without it, I can't proceed and adapt this PR if necessary.
Thanks.

[aagor]

@memsharded Ping.