Time |
Nick |
Message |
09:08 |
|
aabbee joined #evergreen |
09:28 |
csharp |
bshum++ |
11:02 |
pinesol |
News from qatests: Failed Running pgTAP live tests <http://testing.evergreen-ils.org/~live/test.47.html#2019-09-07T11:00:49,903620999-0400 -0> |
11:02 |
pinesol |
News from qatests: Failed Running perl live tests <http://testing.evergreen-ils.org/~live/test.49.html#2019-09-07T11:00:49,931658935-0400 -2> |
12:05 |
bshum |
Hmm |
12:06 |
bshum |
So that PGTap test isn't so bad to fix. We're missing some closing parenthesis in the tests |
12:06 |
bshum |
And one of the test parameters returned a different result than the actual thing being matched |
12:06 |
bshum |
I guess an Inactive patron returns "not_found" rather than "blocked" |
12:07 |
* bshum |
makes a patch to toss on followup |
12:07 |
bshum |
And then I'll see what this perl live test is balking at next :) |
12:22 |
bshum |
Hmm |
12:23 |
bshum |
That live perl test failure seems more complicated |
12:23 |
bshum |
https://bugs.launchpad.net/evergreen/+bug/1817645/comments/1 |
12:23 |
pinesol |
Launchpad bug 1817645 in Evergreen "Configurable patron auth and retrieval" [Wishlist,Fix committed] |
12:24 |
bshum |
In jeffdavis comments on the feature, they described a bug with LWP::Protocol::https that required a newer version from CPAN to resolve, 6.07 and greater |
12:24 |
bshum |
Couple issues... that dependency isn't in the core Makefile.install for the various distros |
12:25 |
bshum |
But also, on Ubuntu 18.04, I have the latest version of that package, and I'm still unable to get successful test results from the API test |
12:25 |
* bshum |
pushes the pgtap fix patch first, and then goes to contemplate lunch |
12:29 |
pinesol |
[evergreen|Ben Shum] LP#1817645: Fix pgtap tests - <http://git.evergreen-ils.org/?p=Evergreen.git;a=commit;h=1421240> |
12:41 |
bshum |
So it's definitely the certificate error affecting the running of the test |
12:41 |
bshum |
Changing the endpoint to http://localhost (without the https) allowed it to succeed |
12:43 |
bshum |
I'd be tempted to let that be the permanent change for the purposes of the test, since the access is restricted to localhost only anyways |
12:43 |
bshum |
and in a real productive environment, someone should be changing it to have a real SSL cert, have real auth user to look up data (rather than the default admin account), etc. |
12:43 |
bshum |
So........... hmm, I'll ponder that |
12:43 |
bshum |
While I eat my sandwich :D |
13:38 |
jeffdavis |
I think the Ubuntu package for LWP::Protocol::https is installed as a dependency when you install libwww-perl, so should be available on any system running OpenSRF. It would make sense for Makefile.install to the later non-buggy version from CPAN though. |
13:54 |
jeffdavis |
Switching to http doesn't feel great but it does still confirm whether the API endpoint is returning correct responses, which is the purpose of the test. |
15:54 |
|
sandbergja joined #evergreen |
18:27 |
bshum |
jeffdavis: I haven't read enough and finished spinning older distro VMs, but if the bug is only from those SSL checks, then switching to http for the test would resolve for everybody's tests |
18:28 |
bshum |
Without needing to upgrade that package |
18:28 |
bshum |
And if I've already got the latest package on 18.04 and it's still broken, that seems like there's something deeper problematic with it than just the version |
18:34 |
bshum |
also systemd-- # apt.systemd.daily install is super annoying |
18:39 |
* bshum |
finishes setting up fresh Ubuntu 16.04 to retest |
19:41 |
bshum |
So much faster to clone virtualbox VMs once they've been compacted down in size, whee! |
20:07 |
bshum |
Very interesting |
20:07 |
bshum |
So, on Ubuntu 16.04, changing it to http for the test, definitely makes it work. So that's consistent |
20:08 |
bshum |
But also installing LWP::Protocol::https via CPAN and restarting apache, also lets the test proceed happily too |
20:08 |
bshum |
I wonder why that's different on Ubuntu 18.04 now, hmm |
21:43 |
bshum |
Reading https://github.com/libwww-perl/LWP-Protocol-https/issues/47 and based on the source for LWP::Protocol::https, I think it's still broken |
21:52 |
bshum |
Seems to me that if we're passing the value to ignore SSL cert anyways, we're already not paying attention to the security as part of the test |
21:52 |
bshum |
So having it go to http instead of https seems moot |
22:27 |
|
sandbergja joined #evergreen |
23:02 |
pinesol |
News from qatests: Failed Running pgTAP live tests <http://testing.evergreen-ils.org/~live/test.47.html#2019-09-07T23:00:56,032084177-0400 -0> |
23:02 |
pinesol |
News from qatests: Failed Running perl live tests <http://testing.evergreen-ils.org/~live/test.49.html#2019-09-07T23:00:56,061217285-0400 -2> |