From mouse Tue Jun 27 00:36:53 2006 Return-Path: Received: (from mouse@localhost) by Sparkle.Rodents.Montreal.QC.CA (8.8.8/8.8.8) id AAA11673; Tue, 27 Jun 2006 00:36:53 -0400 (EDT) Date: Tue, 27 Jun 2006 00:36:53 -0400 (EDT) From: der Mouse Message-Id: <200606270436.AAA11673@Sparkle.Rodents.Montreal.QC.CA> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit X-Erik-Conspiracy: There is no Conspiracy - and if there were I wouldn't be part of it anyway. X-Message-Flag: Microsoft: the company who gave us the botnet zombies. To: mouseware Subject: New moussh moussh is now at 20060626. NOTE: THERE IS A POTENTIAL COMPATABILITY ISSUE WITH THIS VERSION. See the third item below (about the agent extension). New in this version: - INSTALL now documents more accurately (and in more detail) what AF_TIMER sockets are used for, and what happens if they're not available. - An interactive agent interface exists (see the manpage). Interactive agent connections do not work through agent forwarding. - The agent supports an extention with which the agent clients tell the agent what their command lines are. This is primarily for the interactive agent, since that's the only interface by which this information can be gotten back out of the agent. (Note that the agent draft defines the SSH_AGENT_EXTENSION message as 301, even though it's a one-byte field. I use 45, which is 301%256.) The previous version (a) expected agent protocol version 2, while this version does agent protocol version 3, and (b) demanded that the agent response be exactly five bytes (which was correct for version 2). This version's agent client code is compatible with version 2 agents, including older moussh versions, but older moussh client code does not work with agent protocol version 3, such as this version of moussh. (Forwarded agent connections work provided you don't actually *use* the forwarded agent.) - The agent add interface now supports setting constraints (such as expiration time) on keys. Constraints can also be manipulated from the interactive interface. This also means that adding a key which is already present produces a warning, not an error, since with differing constraints it can be useful to have the same key present multiple times. - The agent is now two processes, not one. One of them handles agent protocol, including the interactive interface; the other handles public-key operations. In particular, under the old version, when an agent authentication was in progress, the agent was locked up to the extent that you couldn't even use -list to look at its list of keys. Now, while crypto operations are still single-threaded with respect to other crypto operations, they do not interfere with non-crypto operations or the interactive interface. - I've been writing more internal documentation, in the form of comments. Only a handful of files have nontrivial internal doc, but it's making (slow) progress. - Fixed a bug that produced a panic in the face of certain incorrect calls in the config file. (When (*checkargs)() rejects an arglist, (*deps)() was still getting called to remove dependencies that hadn't yet been added.) - Convert (some of) the lists of manifest constants to enums. This is better style in some respects and makes gdb more effective anyway. - How the list of verbosity constants is generated is reorganized. (This does not affect people who don't hack on the internals.) - Various minor wording improvements in the manpage (in addition to the manpage sides of the above changes). - The way the NESTED* definitions are overridden has changed; see nested.h if you've been using the overrides. - A number of internal interfaces have been reorganized to support the above changes (the stuff in pp.h/pp.c is an example). /~\ The ASCII der Mouse \ / Ribbon Campaign X Against HTML mouse@rodents.montreal.qc.ca / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B From mouse Wed Jun 28 12:40:06 2006 Return-Path: Received: (from mouse@localhost) by Sparkle.Rodents.Montreal.QC.CA (8.8.8/8.8.8) id MAA28868; Wed, 28 Jun 2006 12:40:06 -0400 (EDT) From: der Mouse Message-Id: <200606281640.MAA28868@Sparkle.Rodents.Montreal.QC.CA> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit X-Erik-Conspiracy: There is no Conspiracy - and if there were I wouldn't be part of it anyway. X-Message-Flag: Microsoft: the company who gave us the botnet zombies. Date: Wed, 28 Jun 2006 12:34:09 -0400 (EDT) To: mouseware Subject: YA moussh Oops. When doing my bring-back-into-sync, I forgot to pick up some changes which were emailed to me. I also discovered another crasher bug in the presence of incorrect stuff in the config file, very much like the one I fixed a rev or two ago - I must have missed it when fixing that one. Now at 20060628. Changes: - Update INSTALL (minor wording change one place, talk about a potential build problem another). - Fix a call in one no-AF_TIMER case to pass the correct args. - Remove __transparent__ from some union declarations. (__transparent__ is mine and should not have snuck in; stock gcc supports the same semantics but with a slightly different syntax.) - Don't fall over in client.c if NI_WITHSCOPEID is missing. It seems to have silently disappeared sometime between my 1.4T and 3.0. - Fix the crasher bug I refer to above. /~\ The ASCII der Mouse \ / Ribbon Campaign X Against HTML mouse@rodents.montreal.qc.ca / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B