On  6/03/10 01:09 AM, Liane Praza wrote:
On 03/ 4/10 10:49 PM, James C. McPherson wrote:
On 5/03/10 04:41 PM, James C. McPherson wrote:

Gday one-n-all,
I've put together this script to address

http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6930662
6930662 must provide a tool for gks to update ipkg.sfbay/on-nightly


I've tested it with my repos, and with the nightly gate repo.
All seems to work :-)

I'd appreciate your review comments by 5pm US/Pacific on
Friday 5 March.


and I'd be much better off if I remembered to include the
webrev url! (it's been a long week)


http://cr.opensolaris.org/~jmcp/6930662/webrev/


I've only glanced through the code so far, but wanted to note before I
forget that we should make sure that the bulk of this script goes away
once this bug is fixed:

http://defect.opensolaris.org/bz/show_bug.cgi?id=14965

(The end goal being that our customers who use IPS have to write less
tooling themselves too. You writing this script I think is very
informative for another concrete use case for the productized merge.)

Oh, definitely!


The other note is that because merge currently isn't productized, it
(and your script) are particular about publishing to a specific port.
That's fine, but IIRC if a depotd is open for publication, it allows
anyone to do so. So if the depot startup fails because somebody's
already using that port, you'll end up publishing to someone else's
depot. So either some error checking around failure to start up http:
depots, or a big warning about use of this script would be good.

Yes, that's a rather large liability with it. Frankly, it
really is a hack to get us over the hump, and I know it can
be improved upon.

Ondrej found that the sleep 10 wasn't quite enough, but a 20
second sleep was more likely to work. That annoys me :|

I propose a warning message, something along the lines of


WARNING: If you are running this on a system with other users
likely to (a) also run this script and/or (b) publish their own
repos with pkg.depotd, then you must ensure that this script
uses ports which do not conflict with others. The default ports
are 10002 for x86 repos, and 10001 for sparc repos.




While I'm thinking about it, can we use rsync to transfer
merged repos around systems?



James
--
Senior Software Engineer, Solaris
Sun Microsystems
http://www.jmcp.homeunix.com/blog
_______________________________________________
tools-discuss mailing list
tools-discuss@opensolaris.org

Reply via email to