~go-bot/goose/trunk : changes

~go-bot/goose/trunk » Changes from revision 128

From Revision 121 to 102

Rev	Summary	Authors	Tags	Date
121	[r=gz] Added proper errortype in testservices. [r=gz] Added proper errortype in testservices. Added an error type for testservers that contains a message and an error code. Switched novaservice to use new error type. https://codereview.appspot.com/99960043/ R=axwalk	Horacio Durán		9 years ago
120	[r=gz] tools/secgroup-delete-all: Change test package [r=gz] tools/secgroup-delete-all: Change test package Change tests to be in package main rather than main_test, which may work around an odd issue on the landing bot. https://codereview.appspot.com/91010044/ R=axwalk, hduran-8	Martin Packman		9 years ago
119	[r=axwalk] http: close response body if unused [r=axwalk] http: close response body if unused BinaryRequest calls may discard the response without closing it. This was causing sockets to be left open in juju-core/provider/openstack tests, leading to test failures. https://codereview.appspot.com/90540043/	Andrew Wilkins		10 years ago
118	[r=jameinel],[bug=1226996] panic if we get bad requests [r=jameinel],[bug=1226996] panic if we get bad requests It turns out that if you don't supply CIDR or GroupId Openstack just sets CIDR to 0.0.0.0/0. We don't want to rely on that behavior, so this just changes the goose test double to panic if it got that situation. This can't land until we land my updated Juju branch (or the test suite will appropriately fail with panic().) https://codereview.appspot.com/13781043/	John Arbash Meinel		10 years ago
117	[r=dave-cheney] Make test pass on gccgo 4.8 [r=dave-cheney] Make test pass on gccgo 4.8 gccgo 4.8 and 4.9 have a different view on the number of callers on the call stack. Attempt to detect the correct value automatically.	Dave Cheney		10 years ago
116	[r=gz] Add networks attribute to RunServer params [r=gz] Add networks attribute to RunServer params This enables the starting of a server on a specific network, which is required for some situations where multiple networks exist. This is not well tested at present, to do a proper job some more network related operations need to be added so the tests can do real operations. https://codereview.appspot.com/56670043/ R=natefinch	Martin Packman		10 years ago
115	[r=gz] Add support for network listing api call [r=gz] Add support for network listing api call Lists networks using the os-networks extension, which may backend onto either neutron or nova-network. This should provide a baseline of support even where the neutron api is not available. Goose doesn't yet support detecting if a given extension is supported, that would be nice here as we'll want to know which of neutron, os-networks, and os-tenant-networks are usable. https://codereview.appspot.com/50630044/ R=dimitern	Martin Packman		10 years ago
114	[r=wallyworld] The nova service test double was incorrectly ... [r=wallyworld] The nova service test double was incorrectly ignoring the default security group when starting an instance. It assumed that the default group did not need to be in the list of groups specified when starting, but this is not the case.	Ian Booth		10 years ago
113	[r=dave-cheney],[bug=1254564] Split out implementation speci... [r=dave-cheney],[bug=1254564] Split out implementation specific behaviour Use conditional compilation rather than switching on runtime.Compiler. When the upstream bugs are resolved, we can remove these hacks.	Dave Cheney		10 years ago
112	[r=dave-cheney],[bug=1250253],[bug=1254564] Fix assignment t... [r=dave-cheney],[bug=1250253],[bug=1254564] Fix assignment to private struct member This was detected by the Go 1.2 compiler, but is an issue regardless of compiler or revision. The core problem was assignment to a private struct member in a different package. https://codereview.appspot.com/31890043/	Dave Cheney		10 years ago
111	[r=dave-cheney],[bug=1254564] The currentServiceMethodName c... [r=dave-cheney],[bug=1254564] The currentServiceMethodName changes are foul, but this branch passes all tests with gccgo. You'll need the nil-related changes to have goose work with go 1.2.	Michael Hudson-Doyle		10 years ago
110	[r=wallyworld],[bug=1209003],[bug=1246517] This is to fix bu... [r=wallyworld],[bug=1209003],[bug=1246517] This is to fix bug #1209003 (bug #1246517 duplicate), and bug #1246517 It splits container creation in two parts: i) the container is created with a PUT request ii) the ACL is set with a subsequent POST request. The Swift listing requests in JSON format is addressed as follows: i) append the parameter "format=json" in swift.List. Without this merge 'juju bootstrap' does not work in Havana using ceph-radosgw as a Swift back-end.	Patrick Sodre		10 years ago
109	[r=wallyworld] Do not overwrite http.Client's transport [r=wallyworld] Do not overwrite http.Client's transport Ensure that http.Client's transport has DisableKeepAlives=true without ovwrwriting the transport itself. https://codereview.appspot.com/14840043/	Ian Booth		10 years ago
108	[r=wallyworld],[bug=1239558] Set Close = true on all http re... [r=wallyworld],[bug=1239558] Set Close = true on all http requests. This is needed to avoid this bug in Go's http lib: https://code.google.com/p/go/issues/detail?id=4677 https://codereview.appspot.com/14769043/	Ian Booth		10 years ago
107	[r=jameinel] client.NewNonValidatingPublicClient: new API [r=jameinel] client.NewNonValidatingPublicClient: new API While working on the juju-core side of the code, I discovered another path where we try to connect to the HTTPS server (reading public buckets). So we need a way to disable ssl validation there as well. https://codereview.appspot.com/13396048/	John Arbash Meinel		10 years ago
106	[r=jameinel] test.py: don't update dependencies [r=jameinel] test.py: don't update dependencies There is no reason to update dependencies if we aren't running the juju-core test suite anymore. This is what caused the juju-core test suite to start failing because of the change in mgo that broke socket timeouts. (We landed a goose branch which updated dependencies, which brought in the 'broken' mgo, which then caused juju-core to stop landing code.) https://codereview.appspot.com/13379047/	John Arbash Meinel		10 years ago
105	[r=wallyworld] The previous url tweak caused issues with juj... [r=wallyworld] The previous url tweak caused issues with juju-core because we overload the service type attribute. So revert till the main issue can be addressed.	Ian Booth	juju-1.15.0, juju-1.15.1	10 years ago
104	[r=wallyworld] A small tweak to the juju-tools url in the te... [r=wallyworld] A small tweak to the juju-tools url in the test double to ensure it ends with "/tools"	Ian Booth		10 years ago
103	[r=jameinel],[bug=1202163] client/Client: Add API for non-va... [r=jameinel],[bug=1202163] client/Client: Add API for non-validating SSL This is the final goose step for supporting self-signed certificates for the Openstack Provider (bug #1202163). It adds an API (like was done for the http.Client object) named goose/client.NewNonValidatingClient. It then changes around the code a bit so that we grab the HTTP connection early, and then use it in sendRequest. We also had to change the Authenticators so that we can pass in the http.Client object. Legacy was the only one that used a raw net/http.Client and it should be pretty straightforward how it was changed. Also, to pass in the Client object, I added a new function identity.NewAuthenticator(AuthMode, http.Client). This lets us set private attributes, and allows us to not have external switch statements about how to treat the various auth modes. https://codereview.appspot.com/13360044/	John Arbash Meinel		10 years ago
102	[r=jameinel] http.Client: support InsecureSkipVerify [r=jameinel] http.Client: support InsecureSkipVerify This adds a new API http.NewNonSSValidating which ends up passing InsecureSkipVerify: true all the way down the http stack. Because we were using the global stdlib net/http.DefaultClient to enable connection sharing, etc, we continue to cache the http.Client object. I only create the object once we actually need it, and we only use it for requests that also want to be "insecure". I'm open to suggestions if we want to name the API differently. We could also do it as a flag to New() but that is an API break (meaning juju-core stops building until we patch juju-core). So I felt this was a reasonable method. This is step 2 in my work to support ssl-hostname-verify:false for the Openstack provider. (bug #1202163) https://codereview.appspot.com/13520043/	John Arbash Meinel		10 years ago

Older »