In wit 13 we've been getting an error that looks like the below when we specify a bad revision in our wit-manifest.json
File "/scratch/matthewc/wit/lib/wit/package.py", line 137, in checkout
current_origin = self.repo.get_remote()
AttributeError: 'NoneType' object has no attribute 'get_remote'
This is what the error used to look like
wit.gitrepo.GitError: Command [git log -n1 --format=%ct 531e530646a58dfee78c2c0ad2124800fe36d610] in [/scratch/matthewc/soc-sesame-sifive-ws2/.wit/api-soctemplate-sifive] exited with non-zero exit status [128]
stdout: []
stderr: [fatal: bad object 531e530646a58dfee78c2c0ad2124800fe36d610]
This error message regression was introduced in #222.
In #222 I was thinking that since we've 'downloaded' all the git repo's that we need, that the subdep load will behave the same but without a git clone if we set download=False (and it should have only been a mild optimisation where we don't try to re-bind a package to a git repo each time).
Reading the further downstream logic I don't think this is true, and rather than trying to refactor it right now I feel we should atleast enable the old behaviour
In wit 13 we've been getting an error that looks like the below when we specify a bad revision in our wit-manifest.json
This is what the error used to look like
This error message regression was introduced in #222.
In #222 I was thinking that since we've 'downloaded' all the git repo's that we need, that the
subdepload will behave the same but without agit cloneif we setdownload=False(and it should have only been a mild optimisation where we don't try to re-bind a package to a git repo each time). Reading the further downstream logic I don't think this is true, and rather than trying to refactor it right now I feel we should atleast enable the old behaviour