| Time |
Nick |
Message |
| 08:49 |
|
mmorgan joined #evergreen |
| 09:02 |
|
Dyrcona joined #evergreen |
| 09:15 |
Dyrcona |
Are some tests known to be failing on main at the moment? I'm still getting live check failures with the Angular 21 branch and commits to install the OpenAPI programs. Neither of these should affect Perl live tests. |
| 09:20 |
Dyrcona |
OK. Two tests that failed in my first run this morning passed just now.... Our tests are too rickety. |
| 09:21 |
Dyrcona |
Ok. geosort.t has failed every single time, and it is the only one failing now. |
| 09:23 |
Dyrcona |
Acq tests are passing but I'm seeing things like this for all of them: Use of uninitialized value in string ne at /home/opensrf/Evergreen/Open-ILS/src/perlmods/blib/lib/OpenILS/Application/Acq/EDI.pm line 562. |
| 09:23 |
Dyrcona |
It's not the same every time. |
| 09:23 |
Dyrcona |
This is Debian 13, so maybe we need some changes for Perl 5.40? |
| 09:25 |
Dyrcona |
That geosort test passed when we were doing this a few months ago. |
| 09:26 |
Dyrcona |
I wonder if the issue is the OpenSRF version. I'm using OpenSRF 3.3.3 with main Evergreen. That might not work any more. I'll test again on an Ubuntu 24.04 vm with the same branches. |
| 09:27 |
Dyrcona |
If the tests fail the same, that will rule out Debian 13 as the sole cause. If they succeed, then maybe it is a Debian 13/Perl thing. |
| 10:18 |
Dyrcona |
Is geosort sensitive to timezone? |
| 10:19 |
Dyrcona |
Never mind. |
| 10:20 |
Dyrcona |
Trying to do this all from memory.... |
| 10:29 |
|
collum joined #evergreen |
| 10:35 |
pinesol_dev |
News from commits: LP#2141458: Generate openapi_server from openapi_server.in <http://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=e2cb3e3740d993ab4a15b098526e6db2437635c3> |
| 10:35 |
pinesol_dev |
News from commits: LP#2141458: Add api_ctl and openapi_server to core scripts <http://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=9de6cd3b2b4df8991fd008e133437424b28d704e> |
| 10:43 |
Dyrcona |
32-geosort.t fails on Ubuntu 24.04 with the same error message so it's likely an OpenSRF thing. |
| 10:44 |
pinesol |
News from commits: LP#2141458: Generate openapi_server from openapi_server.in <http://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=e2cb3e3740d993ab4a15b098526e6db2437635c3> |
| 10:44 |
pinesol |
News from commits: LP#2141458: Add api_ctl and openapi_server to core scripts <http://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=9de6cd3b2b4df8991fd008e133437424b28d704e> |
| 10:55 |
|
Christineb joined #evergreen |
| 10:57 |
Dyrcona |
Looks like there is definitely a compatibility issue between OpenSRF 3.3 and Evergreen main. I added it to the Development Meeting agenda. |
| 11:10 |
Dyrcona |
Oh. Maybe it's not what I thought. |
| 11:17 |
Dyrcona |
Think I'll fire up the staff client after lunch to see if this causes further issues. It looks like there are different expectations in Evergreen main and OpenSRF 3.3 about exception messages. I think there's a commit to OpenSRF that we can backport. |
| 11:23 |
|
jihpringle joined #evergreen |
| 11:43 |
Dyrcona |
A'ight, but I can't find it. |
| 11:45 |
Dyrcona |
Hrm... The thing that's failing shouldn't be failing from what I can tell, but it's consistent every time I run livecheck: Can't locate object method "content" via package "OpenSRF::DomainObject::oilsMethodException" at live_t/32-geosort.t line 36. |
| 11:46 |
Dyrcona |
Guess I'll have to try with main and main. |
| 11:51 |
Dyrcona |
The line that blows up hasn't changed since 2020. |
| 11:55 |
mmorgan |
Docker question: I have a clone of the Evergreen repo that I use primarily for docker containers. I've rebuilt the container many times over several days with different branches checked out, but in each container, the patron screens will not load. |
| 11:55 |
mmorgan |
The console says: Failed to instantiate module egPatronApp. I know I can clone a fresh repo, but I'd like to have a better understanding of what the issue might be. |
| 11:56 |
Dyrcona |
Sounds like an Angular issue. Are you using npx ng build instead of ng build? |
| 12:06 |
mmorgan |
I'm using the MOBIUS docker dev container, not doing the ng build explicitly. The puzzling thing is if I build the container pointing to a different cloned repo, the patron screens work fine. |
| 12:07 |
Dyrcona |
What do you mean 'cloned repo?' Are you talking about a clone of the Evergreen repo? |
| 12:09 |
mmorgan |
Yes, a clone of the Evergreen repo stored locally in a different location. |
| 12:11 |
Dyrcona |
Do these repos have the same branch checked out? I'm asking because it looks like a failure to install the Angular staff client. The Evergreen build recently changed in main to build Angular with `npx ng build ...` because we no longer install a global Angular as part of the install process. |
| 12:11 |
Dyrcona |
The Docker image setup script might need changes to take this into account. |
| 12:13 |
Dyrcona |
And, `npx ng build ...` will work even with a global Angular installed. I've been using it for a few months before the patch went in. |
| 12:14 |
Bmagic |
The DEV container hasn't caught up with npg ng build yet. It might be that. I'll check to see if I get the same issue on my end |
| 12:14 |
Bmagic |
npg/npx |
| 12:16 |
* Dyrcona |
remembered rsync, but forgot to install git. |
| 12:17 |
Bmagic |
mmorgan: I'm mistaken, the DEV container has* caught up with this build change. Maybe you're using an out of date DEV container? Run this command to be sure: docker pull mobiusoffice/evergreen-ils:dev |
| 12:17 |
Dyrcona |
It could be some other issue, but that was the first thing that came to mind. |
| 12:18 |
Bmagic |
I might have updated that container and failed to make the announcement on the list. I can't remember when I pushed it if I had an internal debate to spamming the listserv |
| 12:18 |
mmorgan |
On current main, my local repo located at git/evergreen/Evergreen builds and patron screens work fine, but my local repo located at git/evergreen/docker_repo/Evergreen builds but has the patron screen issue. Both are using the same version of the dev container. |
| 12:19 |
Bmagic |
and both folders have git checked out to the same exact commit? |
| 12:20 |
mmorgan |
Yes, same exact commit. Most recently both checked out to current main. |
| 12:21 |
Bmagic |
then the issue must be folder permissions. When you launch the container from the command line, I assume your passing the folder as a mounted volume. I've seen local file/folder permissions wreck havoc |
| 12:22 |
Bmagic |
Generally this issue is on Windows hosts |
| 12:22 |
Bmagic |
just to be sure, run that pull command. I'm curious if it results in a new image |
| 12:23 |
Bmagic |
along thouse lines, tripe check your command for typos and pathing issues to the folder you're wanting to run in the container |
| 12:31 |
mmorgan |
Folder permissions look to be the same. Just did a pull on the dev container (using docker desktop) |
| 12:31 |
* mmorgan |
will rebuild the container |
| 12:34 |
mmorgan |
Here's the command: docker run -it -p 80:80 -p 443:443 -p 210:210 -p 6001:6001 -p 32:22 -p 5433:5432 -v //c/users/michele/git/evergreen/docker_repo/Evergreen:/home/opensrf/repos/Evergreen -h mmm.noblenet.org mobiusoffice/evergreen-ils:dev |
| 12:36 |
Bmagic |
that looks fine to me, is that command issued via powershell? |
| 12:36 |
Bmagic |
Windows has several CLI's. git bash (if installed), WSL (if installed), classic command line, and powershell. In this case, it works best from Powershell |
| 12:37 |
Dyrcona |
Hrm. An OpenSRF test fails on main. |
| 12:37 |
mmorgan |
Yes, powershell. |
| 12:37 |
Bmagic |
I've just confirmed that the DEV container works on evergreen main, no issues with patron interface or any other |
| 12:39 |
Bmagic |
mmorgan: In the scenario's I've seen, the container doesn't actually complete the boot process when the folder perms aren't working. There's a step in the ansible the performs an rsync, which will go by so fast you don't see it. You know it's working if you watch the boot process and it hangs on the rsync for a minute |
| 12:48 |
* mmorgan |
's containers take a while to build :-/ |
| 12:49 |
Bmagic |
the dev one does for sure, because it performs a full install from your provided Evergreen branch. The non-dev containers are much quicker |
| 12:53 |
Dyrcona |
Building vms takes a while also. |
| 13:06 |
mmorgan |
Ooh! Patron screens are working! The only thing I really did was pull a new image of the dev container. Still seems odd, because building from my other repo worked. But I'll take it! |
| 13:07 |
mmorgan |
Dyrcona++ |
| 13:07 |
mmorgan |
Bmagic++ |
| 13:09 |
Bmagic |
mmorgan++ |
| 13:09 |
Bmagic |
you coming to the preconf? |
| 13:13 |
mmorgan |
Bmagic: Yes! |
| 13:13 |
Bmagic |
great, I think what you just went through might be useful for the crowd |
| 13:15 |
mmorgan |
I did panic a bit when a significant part of Evergreen appeared to be broken on main! |
| 13:15 |
Bmagic |
I've been there ;) |
| 13:37 |
Dyrcona |
Well, some of the tests appear to be broken and I can't figure out why. :) |
| 13:37 |
Dyrcona |
:( |
| 13:38 |
Dyrcona |
I find vms and other sorts of images can go bad or be bad from the start. |
| 13:41 |
Dyrcona |
looks like the geosort test is busted on main. |
| 13:56 |
Dyrcona |
Lp 2143582 |
| 13:56 |
pinesol |
Launchpad bug 2143582 in Evergreen "Live test 32-geosort.t failure" [Undecided,New] https://launchpad.net/bugs/2143582 |
| 13:57 |
Dyrcona |
Lp 2143584 |
| 13:57 |
pinesol |
Launchpad bug 2143584 in OpenSRF "check_transport_client test falure on main" [Undecided,New] https://launchpad.net/bugs/2143584 |
| 16:27 |
|
jihpringle joined #evergreen |
| 17:12 |
|
mmorgan left #evergreen |
| 19:07 |
|
jihpringle joined #evergreen |
| 21:07 |
|
sandbergja joined #evergreen |
