VCC: Fix +glob with relative path

[walid-git]

Fixes: #4249

[nigoroll]

Thank you for looking after this issue, but ...

as a general note, I have the impression that more and more PRs now contain super terse commit messages and no additional explanations in the notes either. I can hardly avoid taking this a message towards reviewers in the direction of "my time is more important than yours", and I would really appreciate if we made it a custom (again) to explain which problem you are solving and why you have chosen the particular implementation.

On the actual implementation: Can we not use VFIL_searchpath() with an appropriate vfil_path_func_f()?

[walid-git]

Thank you for looking after this issue, but ...

as a general note, I have the impression that more and more PRs now contain super terse commit messages and no additional explanations in the notes either. I can hardly avoid taking this a message towards reviewers in the direction of "my time is more important than yours", and I would really appreciate if we made it a custom (again) to explain which problem you are solving and why you have chosen the particular implementation.

Apologies for the lack of additional explanation in the commit log, but I though the code was simple enough and self explanatory here. I will try to add some content to it.

On the actual implementation: Can we not use VFIL_searchpath() with an appropriate vfil_path_func_f()?

That wouldn't work, VFIL_searchpath uses access() to test if the file exists then calls func on it if it is the case. In this case, we need to use glob() to test for file existence.

[walid-git]

One question that came to my mind regarding how we handle +glob includes: Should we include all the matching files in all vcl_path directories, or do we only include the matching files from the first directory in vcl_path where we have a match (as we do for regular includes) ?

[nigoroll]

Apologies for the lack of additional explanation in the commit log, but I though the code was simple enough and self explanatory here. I will try to add some content to it.

IMHO, commit messages and ticket notes should not explain the code. They should explain the problem, the solution idea and the reasons for the implementation.

On the actual implementation: Can we not use VFIL_searchpath() with an appropriate vfil_path_func_f()?

That wouldn't work, VFIL_searchpath uses access() to test if the file exists then calls func on it if it is the case. In this case, we need to use glob() to test for file existence.

Right. But would it not make more sense to add an option to VFIL_searchpath to run the glob, rather than adding a different interface with some code duplication?

[nigoroll]

One question that came to my mind regarding how we handle +glob includes

This is a very good point. My first reaction is to include all matches within all path elements, but I am not entirely sure.

[walid-git]

Right. But would it not make more sense to add an option to VFIL_searchpath to run the glob, rather than adding a different interface with some code duplication?

That could be an option. I will give it a try and see how that would look.

This is a very good point. My first reaction is to include all matches within all path elements, but I am not entirely sure.

That is also my opinion, but I need to think more about it.