Expose libgit2 reference database backend for managed code refdb impls.

[nulltoken]

Follow up of #392

[nulltoken]

Doh. Travis doesn't like me. 😿

[nulltoken]

Ooooops.

Output on my box:

------ Test started: Assembly: LibGit2Sharp.Tests.dll ------

Output from LibGit2Sharp.Tests.RefdbBackendFixture.CannotOverwriteExistingInRefdbBackend:
  in Write: referencePtr = 108535680
  in Write: referenceHandle = 108535680

1 passed, 0 failed, 0 skipped, took 49,68 seconds (xUnit.net 1.9.0 build 1566).

Output on Travis:

/home/travis/build/libgit2/libgit2sharp/CI-build.msbuild: error : LibGit2Sharp.Tests.RefdbBackendFixture.CannotOverwriteExistingInRefdbBackend: System.NullReferenceException : Object reference not set to an instance of an object
/home/travis/build/libgit2/libgit2sharp/CI-build.msbuild: error :   at LibGit2Sharp.Core.Utf8Marshaler.FromNative (IntPtr pNativeData) [0x00000] in <filename unknown>:0 
  at LibGit2Sharp.Core.Utf8Marshaler.MarshalNativeToManaged (IntPtr pNativeData) [0x00000] in <filename unknown>:0 
  at (wrapper managed-to-native) LibGit2Sharp.Core.Nativethods:git_reference_name (LibGit2Sharp.Core.Handles.NotOwnedReferenceSafeHandle)
  at LibGit2Sharp.Core.Proxy.git_reference_name (LibGit2Sharp.Core.Handles.NotOwnedReferenceSafeHandle reference) [0x00000] in <filename unknown>:0 
  at LibGit2Sharp.RefdbBackend+BackendEntryPoints.Write (IntPtr backend, IntPtr referencePtr) [0x00000] in <filename unknown>:0 
  at (wrapper native-to-managed) LibGit2Sharp.RefdbBackend/BackendEntryPoints:Write (intptr,intptr)
  at (wrapper managed-to-native) LibGit2Sharp.Core.Nativethods:git_reference_create (LibGit2Sharp.Core.Handles.ReferenceSafeHandle&,LibGit2Sharp.Core.Handles.RepositorySafeHandle,string,LibGit2Sharp.Core.GitOid&,bool)
  at LibGit2Sharp.Core.Proxy.git_reference_create (LibGit2Sharp.Core.Handles.RepositorySafeHandle repo, System.String name, LibGit2Sharp.ObjectId targetId, Boolean allowOverwrite) [0x00000] in <filename unknown>:0 
  at LibGit2Sharp.ReferenceCollection.Add (System.String name, LibGit2Sharp.ObjectId targetId, Boolean allowOverwrite, System.String logMessage) [0x00000] in <filename unknown>:0 
  at LibGit2Sharp.Tests.RefdbBackendFixture.CannotOverwriteExistingInRefdbBackend () [0x00000] in <filename unknown>:0 
  at (wrapper managed-to-native) System.Reflection.MonoMethod:InternalInvoke (System.Reflection.MonoMethod,object,object[],System.Exception&)
  at System.Reflection.MonoMethod.Invoke (System.Object obj, BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] parameters, System.Globalization.CultureInfo culture) [0x00000] in <filename unknown>:0 
            Captured output:
              in Write: referencePtr = 140101970351904
              in Write: referenceHandle = 0

[nulltoken]

Doh. Travis doesn't like me.

Ok. Travis and I are BFF again 😉

@ethomson: Ready for your review, Sir!

[paumayr]

I'm really looking forward to having this RefDB backend in libgit2sharp, but I have a little objection: The libgit2 version of the backend explicitly passes the old_oid / old_target in the write mechanism. This is crucial to ensure that the write only occurs if the previous value is unchanged (optimistic concurrency). Without these parameters I find it difficult to support generic key/value backends (s3, azure storage, etc) for refdb.

Researching a bit further, since this has been idle for a year, I guess the libgit2 interface may have changed in the meanwhile.

Can we update this to reflect the change?

[nulltoken]

I guess the libgit2 interface may have changed in the meanwhile.

@paumayr Indeed, it's evolved a bit. 😉 I'll try and see if I can bring this back up to date in the near future. However, if you feel like giving it a try, please do! We'll be happy to help.

[paumayr]

I'll give it a try!

[nulltoken]

I'll give it a try!

Cool! You may be willing to give a look at the current filesystem based implementation for inspiration.

[alcheByte]

Is this going to get merged?

[nulltoken]

Is this going to get merged?

@alcheByte Not in its current state. See the discussion above 😉

[alcheByte]

@nulltoken sorry, I misread the 'All is well' notification at the end of the thread ;)

I created a fork and resolved the merge conflicts, although I may have botched the handling of the NotOwned/ReferenceSafeHandle. I'm getting a AccessViolationException on the call to NativeMethods.git_repository_free(repo). If I remove the using clause for the Repository instance the exception goes away.

I'm a little confused though. Is the plugged in refdb backend supposed to replace the file system backend or be in addition to it? The .git folder is being created...

[nulltoken]

I created a fork and resolved the merge conflict

@alcheByte Neat!

Although I may have botched the handling of the NotOwned/ReferenceSafeHandle. I'm getting a AccessViolationException on the call to NativeMethods.git_repository_free(repo). If I remove the using clause for the Repository instance the exception goes away.

Could you please create a Pull Request from your fork, it will be easier for us to help you? Just a quick note: I would strongly advise you to create a dedicated branch, rather than working directly on vNext. Then create the Pull Request from this newly created branch.

[alcheByte]

Done, let me know if I botched it, I'm a git/GitHub noob :)

So is the plugged in refdb supposed to be in lieu of the fs backend?
I wasn't expecting the .git folder to be created when the refdb has been set...

[nulltoken]

@alcheByte I've taken a quick look. I think the issue you're encountering is related to the fact that the GitRefDbBackend doesn't match the libgit2 interface.

[nulltoken]

Hmmm. Let's move this over to #797