Brock Pytlik wrote:
Stephen Hahn wrote:
* Bart Smaalders <[email protected]> [2009-07-01 21:36]:
http://cr.opensolaris.org/~barts/9808/
9808 subcommand to output manifest useable by include/publish desired
pkgsend.1.txt:
"generate" seems to be unnecessary if we had a dry-run capability in
pkgsend(1). (I believe it would fall out automatically with a
client-side mode.) Why do we need a bundle analyzer again? (I read
the bug report, but there's no explicit use case there. How would
this feature be used? (If everyone knows but me, then please just add
that use case as an example in the manual page...))
Unless I'm very confused, the reason we want a bundle analyzer is to
make initial generation of a manifest easy for a user. For example, a
packager can take their proto area and point pkgsend generate (or dry
run, I don't have strong feelings about the command line) at it and get
a manifest they can use as a basis to add dependencies to, tweak or
change, and then do the final pkgsend. The same logic applies to a tar
bar or System V package that the packager is using a base for the new
package they're publishing.
As a side note, I think the reason a subcommand was suggested was that
it was possible there might be other options we wanted to add to the
command (though I'll admit I can't think of one right at the moment).
I believe you can do a 'dry-run' sort of today by specifying a null://
URI for the destination package repository.
However, that transport class would have to be expanded to optionally
dump the resulting transaction to a file or at least keep a buffered
string internally that could then be printed to stdout.
Cheers,
--
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
