On 10/25/11 17:26, Danek Duvall wrote:
Brock Pytlik wrote:
https://cr.opensolaris.org/action/browse/pkg/bpytlik/18012-v1
Doesn't this constitute a (compatible) API version change?
Oops, good point.
image.py:
- line 2432: is a user ever likely to get into a situation where he's
told that a manifest can't be found, and yet he can see it exists? If
so, perhaps the exception raised here should allow the UI to indicate
that the manifest does not properly verify.
So you're worried about the situation where a manifest on disk is bad,
we tell transport to get the manifest, which fails, so we report a
transport error to the user, and then the user looks in
/var/pkg/publisher/<mumble> and sees that a manifest file is in fact there?
If you want, I guess I can add such a message but I guess I don't see
much value in it since I suspect the number of users looking inside
/var/pkg/publisher/<mumble> is pretty small. At that point, the user's
problem isn't that the manifest on disk is corrupt, it's that they can't
get a correct version of the manifest.
tests:
- You could go for more consolidation in the testsuite if pkgsign()
generated sign_args automatically. You repeat what seems to be the
same clause over and over again for generic "please sign this"
functionality.
To me there's enough variation that it's not worth commonizing, but I'll
take another look and see what I can do.
t_pkg_info.py:
- On a related note, in t_pkg_info.py, you modify the manifest by adding
a license action to it. Why do that instead of signing it? Or adding
a simpler action, like set, that doesn't have a payload?
If I remember right, it's because I wanted to add an action which had an
effect on the output of pkg info. So we do the pkg info locally, and
make sure that the license isn't on the manifest, but that the license
is on the manifest when we do -r.
t_pkg_install.py:
- I'm not understanding how the code in test_bug_18012_keep_installed()
does what the comment says. Or, for that matter, has anything to do
with bug 18012.
Well, first we publish an unsigned package and install it. So on the
system, the package isn't signed.
Now, the package gets signed in the repository but obviously not in the
image.
Then pkg verify is called. If it replaced the manifest on disk with the
updated (now signed) package, then verify would pass because the package
would be signed. Since the verify fails, then we know that verify used
the old, unsigned, package to verify the image.
It has to do with bug 18012 because we need to make sure that verify
doesn't replace the manifest of an installed package, even if the
installed package's manifest is corrupt.
t_pkg_verify.py:
- line 185: "runing" -> "running"
Thanks.
Brock
Thanks,
Danek
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
