chore: Up-port from develop-2.0.0 from week 32/33 - #4141
Conversation
Sorry, something went wrong.
* fix: GC allocations * Add message receive allocations test * Revert the fix to check ILPP build fails the test * Revert forced failure * Remove gc checks on NetworkListTests
* Re-enabled ps5 and webgl * ci: Add WebGL PlayMode test job (MTT-15569) Split the WebGL CI into a build phase and a run phase. webgl-build.yml now exposes the built player as a dedicated 'players' artifact, and a new webgl-test.yml run job consumes it and executes the PlayMode tests inside Firefox on a GPU-backed Ubuntu agent (Unity::VM::GPU, rtx2080). The Firefox browser flags are passed explicitly so UTR does not attempt to download the browser from Stevedore. The run job is wired into the Nightly and Weekly (QV) triggers only, not into PR triggers, since WebGL failures are infrequent and the build is slow. Runtime RuntimePlatform.WebGLPlayer exclusions are intentionally deferred to be added reactively after the first real CI run. * corrected images * disabled webgl test * adjustments * Removed WebGL changes * reverted change * added diagnostic for switch * corrected Switch sdk version * temporarily disabled switch * corrected ticket description
Changde users exclusion to pattern exclusion in pr description check
* Updated changelog and package version for Netcode in anticipation of v2.13.1 release * typo in build automation * trunk shadow changes: com.unity.netcode.gameobjects [skip ci] * Revert CHANGELOG changes * Added UNITY_TEST_FRAMEWORK_1_7_OR_NEWER guard and UnityCoreClrExplicitDisabledAttributeShim.cs to make attribute usage possible * corrected svc bot exclusion --------- Co-authored-by: netcode-automation <svc-netcode-sdk@unity3d.com> Co-authored-by: Michał Chrobot <michal.chrobot@unity3d.com> Co-authored-by: PETS automation <299490404+pets-svc[bot]@users.noreply.github.com> Co-authored-by: Emma <emma.mcmillan@unity3d.com>
Co-authored-by: unity-renovate[bot] <120015202+unity-renovate[bot]@users.noreply.github.com>
|
|
Sorry, something went wrong.
There was a problem hiding this comment.
💡 Harness Review
The allocation-focused runtime changes are generally coherent, but the new integration coverage is not excluded from a known failing CoreCLR configuration and two diagnostic/validation changes alter behavior incorrectly.
Reviewed commit c30cbc3
🤖 Helpful? 👍/👎
Sorry, something went wrong.
|
|
||
| [TestFixture(HostOrServer.Host)] | ||
| [TestFixture(HostOrServer.DAHost)] | ||
| internal class MessageReceiveAllocationTests : NetcodeIntegrationTest |
There was a problem hiding this comment.
This new Host/DAHost integration test creates a second network-manager instance (NumberOfClients is 1) and exercises RPC and NetworkVariable traffic, but it is not marked UnityCoreClrExplicitDisabled. The same change marks comparable multi-instance tests with UUM-149591, whose stated failure mode is that CoreCLR sessions cannot start/connect or time out. Consequently CoreCLR runs will schedule this test and fail or time out instead of skipping it. Add the matching exclusion to the class (or this test) until that issue is fixed.
🤖 Helpful? 👍/👎
Sorry, something went wrong.
| return $"{entry.Key} | {entry.Value.Method.Name} | {invokePermission}"; | ||
| })); | ||
| var invokePermission = permission; | ||
| networkManager.Log.Info(new Context(LogLevel.Developer, "RPC Table Contents").AddCollection(rpcsForBehaviour, entry => $"{entry.Key} | {entry.Value.Method.Name} | {invokePermission}")); |
There was a problem hiding this comment.
invokePermission is the permission of the single RPC that threw, but this callback is invoked once for every entry in rpcsForBehaviour. For a behaviour with mixed RPC permissions, an exception now produces a table claiming every RPC has the failing method's permission, making the diagnostic misleading precisely when it is needed. Look up permissionsTable[entry.Key] inside the callback, as the previous code did.
🤖 Helpful? 👍/👎
Sorry, something went wrong.
| 'svc-' | ||
| const skipUserPatterns = [ | ||
| /^unity-renovate/, | ||
| /svc/ |
There was a problem hiding this comment.
This unanchored pattern bypasses description validation for any login containing svc, not just service accounts—for example, a normal author named alexsvcdev. That silently defeats the required-section gate for those PRs. Keep the intended service-account prefix semantics (such as /^svc-/) or list the specific accounts.
🤖 Helpful? 👍/👎
Sorry, something went wrong.
Reviewers
At least 1 approving review is required to merge this pull request.
Purpose of this PR
PR that up-ports PRs that landed during last 2 weeks or otherwise since we did the last up-port.
Note that I won't "squash and merge" and I will rather merge them separately as then in the future it will be easier to reason about PRs history.
Also not all commits are ported because some of them were up-ported already (picture shows the range of commits I was evaluating)
Jira ticket
N/
Documentation
N/A
Testing & QA (How your changes can be verified during release Playtest)
Green CI
Up-port
This is an up-port
Backports
N/A