Refactoring: Remove usage of deprecated LocalDirs field in SolveOpts

[leandrosansilva]

As client.SolveOpt.LocalDirs is marked as deprecated, this PR moves all usage to LocalMounts instead.

Unfortunately the PR ended up quite long, but the changes are essentially mechanical.

I am afraid that this PR will also conflict with existing open PRs that touch tests, but I believe that resolving the conflicts will be trivial.

Once this PR is merged, client.SolveOpt.LocalDirs will be ready for removal, in case no external projects is using it.

[leandrosansilva]

Note to the reviewer: I did not manage to get the integration tests to run reliably on my setup, so I hope having the tests running on github will be able to reveal issues.

To be honest, the tests don't run reliably for me even on master.

[jedevc]

Thanks @leandrosansilva 🎉

Could we have one usage of LocalDirs that checks that the deprecated property still works? Probably makes sense as a new test in client_test.go.

Otherwise looks pretty good to me (assuming tests pass 👀). One idea I remember having in #4094 was to use our StaticFS implementation in this PR to avoid needing to create temp directories at all:

for example, we could use our Static filesystem implementation in tests to prevent creating lots of temporary directories, or we could use our Merge filesystem implementation to allow easily creating variants of a single context.

If that seems interesting to you, feel free to try that out, but definitely not a blocker 🎉

[leandrosansilva]

Thanks @leandrosansilva 🎉

Could we have one usage of LocalDirs that checks that the deprecated property still works? Probably makes sense as a new test in client_test.go.

Good idea. Done.

Otherwise looks pretty good to me (assuming tests pass 👀). One idea I remember having in #4094 was to use our StaticFS implementation in this PR to avoid needing to create temp directories at all:

Interesting, I will have a look at that.

for example, we could use our Static filesystem implementation in tests to prevent creating lots of temporary directories, or we could use our Merge filesystem implementation to allow easily creating variants of a single context.

If that seems interesting to you, feel free to try that out, but definitely not a blocker 🎉

Great! One test got broken by my changes, but it's now fixed and pushed. Again, I could not get all the integration tests to pass locally, but I could fix the test that the previous execution on github pointed out.

[leandrosansilva]

My apologies, I ended up committing some code I should not have committed. It's weird that running validate-lint locally did not find the issue. Fixing my mess now.

[leandrosansilva]

Hi @jedevc, FYI the PR is ready for review again, so please feel free to do so whenever suits you.

I will have a look at the StaticFS story, but I believe it's better done in a further PR.

[jedevc]

Thanks @leandrosansilva! Much appreciated!