> Thanks for fixing it.
>
> Is there a particular patch this should be squashed into, or does it
> stand alone? It the latter, mind writing a commit message for it?
Not sure if I'm using "fixup" correctly in the subject, but this is
meant to be squashed onto the tip of jt/http-auth-proto-v2-fix -
specifically, deb7d2094a ("remote-curl: use post_rpc() for protocol v2
also", 2019-02-14).
I don't think it should be a standalone commit, as the tip is buggy -
which might break bisect. But if we really want a standalone commit,
this commit message should work:
remote-curl: handle consecutive flushes correctly
When the client, using protocol v2, sends two consecutive flushes and
then an EOF, remote-curl dies. This is because, at the start of a
new request, stateless_connect() reads, and if no EOF is found, then
stateless_connect() calls post_rpc() which reads until a flush is
encountered. This is a problem when stateless_connect() reads the second
consecutive flush (hence, no EOF), and then post_rpc() reads, not
expecting an EOF at all.
Teach stateless_connect() to inform post_rpc() to read only if what it
read isn't a flush.
