Ian Jackson wrote: > This is very odd. The error message is misleading. What it's > actually doing is looking at .git/refs/remotes/dgit/dgit/sid, and > finding either an empty line or ENOENT. I don't understand how this > can fail like this if what you say is right. > > Can you please double-check for me the results of > cat .git/refs/remotes/dgit/dgit/sid
Hmm. I should have snapshotted the tree. Fail. In my hurry to get a release out, I ended up manually rming that file (or trying to -- not sure if it was there). I also manually edited .git/packed-refs and I know I deleted a ref for dgit/sgit/sid out of there. [I did actually happen to cp -a that tree for other reasons at around this time, but I am not sure at exactly what point I took this copy and so I cannot rely on this being accurate. It's probably from a point in time slightly before the dgit push ran. joey@darkstar:~/tmp/git-annex>cat .git/refs/remotes/dgit/dgit/sid 08f7dc87a5da1173184d39ec93b80233e661973e joey@darkstar:~/tmp/git-annex>grep dgit/dgit/sid .git/packed-refs - exit 1 ] Is dgit aware of the packed-refs file? I did not quickly find any code in it that looks in .git/refs/ manually, but if it does it should really be using git show-ref instead. -- see shy jo
signature.asc
Description: Digital signature