* Danek Duvall <[EMAIL PROTECTED]> [2008-06-17 16:40]:
> On Mon, Jun 16, 2008 at 02:04:32PM -0700, [EMAIL PROTECTED] wrote:
> > * Why did you add the additional call to symlink in the install_file
> > logic? A comment explaining this would be nice.
>
> That's part of Stephen's code.
The patch I sent to Danek was a prototype of coalescing the package
states into a single $META/state/installed directory. In the patch,
each installed package gets a symlink in $META/state/installed to its
$META/pkg/[stem]/[version]/installed state file. The various
gen_installed_pkgs() simplify to an iteration over a directory of all
positive entries, rather than a descent through a tree of negative and
positive entries.
I would expect to see the $META/state/ directory to be expanded to
include at least $META/state/transit-install, for packages undergoing
installation. (These would be consulted during the evaluation of
package removals that might be dependencies of an ongoing install.)
There are a couple of questions:
- should the symbolic link point to the installed file, or to the
$META/pkg/[stem]/[version] directory, and
- should we use a link at all, or does our Windows portability require
use of an empty file? (Or should the original installed file just
be moved?)
There's also an implementation question about keeping image metadata
versioning code organized by module, or having one or more
update-metadata modules. The current patch uses the former, and makes
the update on the first modifying operation.
I'll probably have to hand this patch off anyway, since I'll be out of
the office for a bit the next couple of weeks.
- Stephen
--
[EMAIL PROTECTED] http://blogs.sun.com/sch/
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
