On Fri, 27 Jun 2008, Will Fiveash wrote: > On Fri, Jun 27, 2008 at 09:46:11AM -0400, James Carlson wrote: >> Will Fiveash writes: >>> I'm cc'ing scm-migration-dev at opensolaris.org on this. >>> >>> On Thu, Jun 26, 2008 at 05:18:07PM -0500, Will Fiveash wrote: >>>> I recently added a shell script (krb-diag) to the opensolaris kerberos >>>> hg repository: ssh://anon at hg.opensolaris.org/hg/kerberos/tools >>>> >>>> What confuses me is that I looked at the scripts in the onnv-scm hg >>>> repository (onnv-scm/usr/src/tools/scripts) and I see unexpanded SCCS >>>> keywords in the script files. So I left the unexpanded SCCS keywords in >>>> krb-diag. I'm betting this was wrong since the script should have a >>>> current version number in it. What was I supposed to do here? >>> >>> Is this documented somewhere on >>> http://www.opensolaris.org/os/community/tools/scm/hg_teamware_transition/ >>> ? >> >> You did the right thing. >> >> All of the old SCCS keywords will be ripped out after the gate itself >> converts to Mercurial. They don't do anything, and there's no >> reasonable way to insert anything equivalent for Mercurial. (Hg >> doesn't have per-file version numbers or expanding keywords like SCCS >> or RCS.) > > The problem I face is that krb-diag outputs its version which is useful > when looking at its diagnostic output. As it is now in the kerberos > repository the unexpanded SCCS keywords will be output which is useless. > In addition, the idea is that the version of the script in the krb > repository will be the canonical version and people should be able to > download that version and run it (with the version info embedded in the > script). Will this require a Makefile to produce a version of the > script with its version info expanded?
ON recently has added a new rule that no SCCS keywords can be used as version strings going forward. There have been a few suggestions from some SCM migration folks on alternatives, but most won't really be exactly the same as what we have now & won't work until we move to hg. ( http://opensolaris.org/os/community/on/crt/rti-nits/ ) for example, using a macro in a makefile that uses "hg tip" that can be used for version information. >> I didn't find that in the transition document, and I'm not sure why. >> The folks writing those documents have used hg for a while, and >> perhaps it just seemed too "obvious" -- though it's not for someone >> with a Teamware background. It probably should be there. > > I agree. This is one of the first things I needed help with. There was a KTD given by Mike Kupfer & Stephen Lau in 2007 on this, that covered this as well. It's available internally on http://ktd.eng/ , but as far as I know the slides are not also available externally. (perhaps someone cc'ed here from the SCM team could do so?) > >> For now, until the transition, leave them in place as they are, and do >> *NOT* build any code that relies on those strings. (Such code was >> almost invariably kludgey anyway.) >> Actually, Jim, our recommendation is to start removing them now. The SCM team filed many bugs, but unfortunately didn't find all of the instances of SCCS IDs being used this way. The CRT (Change Request Team) is now keeping their eyes peeled for these and trying to get them fixed before the transition. The SCM migration team will not be, to the best of my knowledge, fixing all of these instances themselves - that's why they filed bugs, to help share the load (and let the subject matter experts decide how they wanted to handle version strings going forward and fix any kludgey code) Valerie -- Valerie Fenwick, http://blogs.sun.com/bubbva Solaris Security Technologies, Developer, Sun Microsystems, Inc. 17 Network Circle, Menlo Park, CA, 94025. 650-786-0461
