On 11/21/11 11:53, Brock Pytlik wrote:
On 11/21/11 11:34, Shawn Walker wrote:
On 11/18/11 17:56, Brock Pytlik wrote:
On 11/11/11 14:02, Shawn Walker wrote:
Greetings,
The following webrev contains fixes for the following issue:
18716 captive_portal_test requires at least one configured source to
be reachable even when specifying temporary sources
https://cr.opensolaris.org/action/browse/pkg/swalker/pkg-18716/webrev/
This makes it possible to install from temporary sources (such as
package archives) even when all other sources are unreachable with two
caveats:
1) you must use --no-refresh with install/update -- that's bug 18323;
not this bug
2) obviously all of the package data required for the operation must
be provided by one or more of the temporary sources specified
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
transport.py:
This code all seems to be making the assumption that there's a single
alt_repo. Perhaps I just don't have the entire call stack set up in my
head, but info and other commands can take multiple -g arguments.
(Because of the difference syntax that search uses, I think it can only
take a single alt repo.) It seems like if you had no configured
publishers and two -g's specified on the command line, one an available
on disk archive and one an unavailable network repo, we might pass the
captive portal test (depending on which alternate repo was used for the
captive portal test).
Transport can only use one temporary (alt_repo) source at a time, so
callers of transport are responsible for designating which of those
sources to use for the duration of the operation.
Ok, so in the case of multiple -g options, which is chosen to be used
for the captive portal test?
Whatever the first source is to be contacted is the one that is used.
So we don't necessarily use any of the temporary sources for the captive
portal test.
t_pkg_temp_sources:
I'm confused as to why setting the signature-policy over and over is
necessary.
It's only setting it once per image. Note that every subsection of the
test creates a new image.
Ok, then let me rephrase. I'm confused why it's necessary to set at all.
Because the tests would currently fail since the test isn't setup with
the certs available. That, and I didn't see the point of testing cert
functionality here, it just makes testing slower and we test cert
functionality somewhere else.
As described above, I'm a little concerned about using multiple -g's,
can you include some tests for these situations? (For example, 2
alternate repos are specified and both are available, 2 are specified
but only one's available...)
If you specify two alternate sources and one is not available, the
operation will fail. All temporary sources you specify must be
available. That is intentional.
I agree, the operation should fail. The question for me is whether it
fails during the captive portal test or not. I would still like to see
some tests here if we don't already have them.
As we discussed offline, I'll file a bug for captive portal test as it
has several issues we discussed.
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
