Han-Wen Nienhuys

On Wed, Mar 25, 2020 at 3:09 PM Patrick Steinhardt wrote > > - > > At some places I feel like there's high potential for namespacing > issues. E.g....

For the ref-like objects, this is what JGit does https://eclipse.googlesource.com/jgit/jgit/+/459fc28cf6a22992b7d168cfd52286cd76ca68bc/org.eclipse.jgit/src/org/eclipse/jgit/lib/RefDatabase.java#511 (https://eclipse.googlesource.com/jgit/jgit/+/459fc28cf6a22992b7d168cfd52286cd76ca68bc/org.eclipse.jgit/src/org/eclipse/jgit/internal/storage/file/RefDirectory.java#113) In JGit, HEAD is expected to be managed by RefDatabase. Unfortunately, nobody has been able to articulate to me...

Hey Patrick, can you check out https://godoc.org/github.com/google/reftable#Addition and https://godoc.org/github.com/google/reftable#Stack.NewAddition would that work as an API for you?

https://github.com/google/reftable/commit/7aabfe0f39ce40f47a9a0f63e9c8b10f753e5afe for the transaction api in C.

BTW, - I would be grateful for any feedback about reftable on windows. I don't have windows, and I'd like to keep it that way.

@pks-t - does libgit2 already have code to detect file / directory conflicts? Looks like I may need to write and/or refactor that to check transactions, and wondering if that...

ok. I'm working on refname verification. I finished the Go code https://github.com/google/reftable/commit/2abf6f32a0f17e720d40e821ea224ac3e7618760 , and will try to get the C code done today.

refname validation will be done automatically if you upgrade to the latest version: https://github.com/google/reftable/commit/c8f1f44b463ca1ddd96efebe1b009b00873bf628

> @hanwen I've found one bug (or at least unexpected behaviour) with the reftable implementation: if I call `reftable_backend_write_records` with no reference records but `n` log records, where each of...

ref records are keyed by ref_name. log records are keyed by (ref_name + update_index). When you write a normal update of a ref, you should write the ref record and...