Bugfix/user in share20ocscontroller#34568
Conversation
|
fbf8f28 to
2b6a07d
Compare
Codecov Report
@@ Coverage Diff @@
## master #34568 +/- ##
=============================================
- Coverage 65.16% 48.43% -16.73%
=============================================
Files 1200 109 -1091
Lines 69684 10504 -59180
Branches 1283 1283
=============================================
- Hits 45407 5088 -40319
+ Misses 23903 5042 -18861
Partials 374 374
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## master #34568 +/- ##
============================================
- Coverage 65.22% 65.22% -0.01%
- Complexity 18434 18438 +4
============================================
Files 1203 1203
Lines 69816 69827 +11
Branches 1280 1280
============================================
+ Hits 45540 45544 +4
- Misses 23904 23911 +7
Partials 372 372
Continue to review full report at Codecov.
|
2b6a07d to
56b135b
Compare
|
@DeepDiver1975 All the failing federation scenarios are when something is shared on the "local" server and the receiving user on the "remote" federated server does something in the share that they have received (uploads a file... to the share). The action does not seem to be effective - i.e. the result is not seen on the "local" server. In the drone setup, the "local" server is running the code of the PR branch. The "remote" federated server is running the code in the last QA tarball (i.e. recent code, but not the PR). That helps a little bit to know that the a new change will work in a federated environment with old versions. Perhaps the PR code is causing some change in the way that incoming "old style" requests from the "remote" server are not working/valid on the "new style" local server. Or perhaps there is some issue with the test runner code (which matches the new PR) correctly accessing the API on the "old" remote server. Post if you make progress. I will look locally when I get to the office in the morning. |
|
Rebase after PRs #34577 and #34582 are merged to master. That will provide nicer handling of the federation scenarios. The issue seems to be around: When the remote federated server gets an API request to accept a pending remote share, it sends off to the local server. The local server gets a 500, but the remote server does not pass that on. The remote server responds t its API request with an HTTP 200 an no OCS status. |
@phil-davis Can I ask you to help out here to find the log messages for the 500 status code? I'm a bit lost 😕 |
0880f94 to
4e69c57
Compare
|
Assigned myself also so I notice this on Monday. |
@phil-davis any news? Is there anything I can do? THX |
4e69c57 to
2418390
Compare
|
Here is the next clue ;) After running this once, it passes but on the remote federated server:
When I run the scenario again, it fails because the scenario creates a new federated share, and now there are 2 of them: The scenario tries to accept that stale external share entry and upload-overwrite When checkout I think the underlying problem is that On the local server, when the scenario is finishing and deleting the users, I see: It seems that, as it is deleting a user, the federated server sends a When running the scenario from |
|
Maybe another clue: But from master: The test scenario cleanup deletes the local users first, then the remote users. So the local user (and things that the local user shared) do not exist any more. Then the remote user is deleted, and the above kind of thing happens - that |
this is wrong - a proper OCS response has to be sent and has to be accepted. What is the response on stable10? Mind checking this @phil-davis ? THX |
|
I tried |
|
@DeepDiver1975 see PR #34622 that is on top of this. It makes the response be You might like that code, or might want to just return a |
|
is this related #34626 ? |
|
one more #34636 |
|
closed in favor of #34622 |
|
The code here will get backported as part of #34622 |
Description
Before:
After:
Related Issue
Motivation and Context
Don't break behavior!
How Has This Been Tested?
Types of changes
Checklist:
Open tasks: