Date: Thu, 20 Nov 1997 17:59:54 -0700 From: Nate Williams <nate@mt.sri.com> To: chad@dcfinc.com Cc: nate@mt.sri.com (Nate Williams), rkw@dataplex.net, freebsd-stable@freebsd.org Subject: Re: Version Resolution? Message-ID: <199711210059.RAA12303@mt.sri.com> In-Reply-To: <199711210046.RAA04475@freebie.dcfinc.com> References: <199711210040.RAA12221@mt.sri.com> <199711210046.RAA04475@freebie.dcfinc.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Chad R. Larson writes: > > > But there are (can be) seperate copies of newvers.sh in each branch? > > > Seems it would be an administrative procedure to touch it correctly as > > > part of creating a new branch. How often does that happen? Couple of > > > times a year? > > > > More than that, but as I stated before, file corruption even *once* > > isn't acceptable. The number of problems that will occur when a branch > > occurs (and everyone and his dog decides to finally download things) > > means that the # of questions due to bogus files is now bigger than the > > number of questions solved by implementing the solution. > > I'm obviously ignorant of the discussions you had with Richard. Can you > explain the "file corruption" you imagine? *sigh* OK, we have a file. It's got a couple branches in it (branched at version 1.6 (BRANCH1) and at 1.30 (BRANCH2)). The current 'HEAD' revision is 1.101. So, in the file we need to have the following: 1.6.1.1 is the first revision on BRANCH1, the next revision is 1.6.1.2, etc.... 1.30.4.1 would be the first revision on BRANCH2 (Don't ask why the '4' is there, since I don't know. :) Each subsequent revision on that branch would have a number 1.30.4.2, 1.30.4.3, etc.... On the HEAD, we're at 1.101. Let's say a new branch is created on the HEAD. At that point, the RCS file would contain the same as above, *PLUS* a new set of branch revisions, say 1.101.6.1: However, the 'auto-RCS file' generator would spam this file and kill it (so the new branch would have no 'tag' file in it), so CVSup would have to know how to 'remove' the valid file, one that it may have sent just 5 minutes ago to the same client (assuming the client connected right at the wrong time.) So, we've got a file that has this new branch in it, and 5 minutes later the file is totally different, thus confusing CVSup and any CVS clients assuming that CVSup/CTM actually end up finally getting the file straightended out. (My client may have thought that the 1.106 branch was valid, and then that branch tag was ripped out from underneath it, causing problems.) Now, the possibility is small, and only within the window of someone going and fixing the 'auto RCS-generator' file, but given the hecticness of releases and the simple fact that 'magic' things never get done, having it done automatically is much more important than requiring someone to go edit a file and make sure things don't get spammed in the process. Especially when it *can* be done automatically. Nate
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199711210059.RAA12303>