Re: error: error in sideband demultiplexer
- From
Junio C Hamano <gitster@pobox.com>
- Date
- Jun 14, 2011, 19:32 UTC
- Message-ID
- <7vhb7sky4i.fsf@alter.siamese.dyndns.org>
- In-Reply-To
- <20110614164943.GA26764@sigill.intra.peff.net>
Jeff King <peff@peff.net> writes:
Show 26 quoted lines
> On Tue, Jun 14, 2011 at 01:20:32PM +0100, Richard Shaw wrote: > >> Could someone please give me an explanation of what the following error message means: >> >> error: error in sideband demultiplexer > > The git protocol generally operates over a single stream. Most of that > stream will have object data on it, but we also want to be able to pass > informational messages. So there is a "sideband", which works something > like: > > 1. The remote end puts a header at the beginning of each packet > telling us whether it's for the data stream or the sideband. > > 2. We either fork or start a thread to read the data from the remote > and demultiplex it. If it's sideband, we output it to stderr. If > it's data, we pass it on to the main git program. > > The error message indicates that the thread or forked process > implementing the sideband demultiplexer returned an error in its exit > code. Usually it will have printed some other error message already that > is more specific, but you may have found a case where it doesn't. > > Which platform are you on? Are you building with threads (i.e., are you > setting NO_PTHREADS)? Are there any other error messages? If not, is it > possible to use "strace -f" to get a trace that shows the error?
Another common question to ask is...
Do you have something funky in dot files under $HOME on the remote side (e.g. .bash_profile or .login) that cause the process started by the non-interactive ssh session we make to the remote end give extra output?
IOW,
$ ssh that-remote-site "cat /dev/null"
should not give us any extra "greeting".