You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Our current V3 release is at version 3.22.0, with a single bug fix merged but unreleased. Adding a .Net 11.0 agent would, of course, call for a 3.23.0 release.
I think that the 4.0 release is probably a few months away at best, so this seems like a decent idea. It's not, however one I want to work on.
So this issue, while a Yes / No decision for the @nunit/engine-team , comes with two varieties of "Yes"...
Yes, you can assign it to me!
Yes, I can't do it but I hope somebody else will.
If somebody takes this on, we'll roll that bug fix into the 3.23 release. Otherwise I'll do a 3.22.1 release.
@nunit/engine-team Related to this, I'm retiring the "V4" label but keeping "V3." All open issues not labeled "V3" may be considered as V4 issues in future.
@CharliePoole, I support adding a .NET 11 agent if someone can take responsibility for the implementation and release. With V4 still some months away, that would benefit existing V3 Console users.
Our current commitments remain the V4 remote-provider work and adapter integration. I can make our existing .NET 11 fixtures and validation results available to whoever takes this on, but I’m not volunteering for ongoing V3 release maintenance.
Prepared by Neil Colvin with assistance from OpenAI Codex.
I don't think any of our work so far on V4 is needed. Agents in v3 are built-in and are simply different runtime targets of a single assembly. Creating one is a matter of adding net11.0 as a target for the agent and its tests and updating the workflow. Making it work is a bit more, but I think pretty trivial. As stated earlier, I suspect net11.0 will issue even more warnings than net10.0, but that's not hard to fix either.
But every new agent I've added triggers more bug reports and I don't doubt this will too. That's the big job for the person who is going to volunteer. :-)
I think we should do it - and I can take a look at doing this (possibly this weekend). I cannot guarantee that I'll have time for all the bug reports :-)
@mikkelbu, I can share our existing .NET 11 fixtures and expected outcomes for validation of the V3 built-in agent.
One separate V3 compatibility fix to keep in view is draft #1955: it adapts the modern driver to NUnit 5’s parameterless StopRun API while preserving the older Boolean signature. It does not add a .NET 11 agent or depend on that work, but could be considered for the next V3 release once reviewed.
Our main implementation commitments remain V4 remote execution and adapter integration.
Prepared by Neil Colvin with assistance from OpenAI Codex.
I've changed the title and labeling of this issue and assigned it to you. The version3 branch already contains fixes for #1927 and #1856. I'm re-targetting those to the 3.23 release so they'll appear in the release notes. I'll send a follow-up note to you about how to do the actual release.
@OsirisTerje Please check what's in version3 head to make sure you have everything you want for the adapter.
@CharliePoole Unfortunately not. Last weekend "disappeared" due to illness in my family (nothing major, but just took my time) and since then I've been (too) busy with work. I hope I've some time today, but next is also going to be busy (at least the working days)
I only had time to make the obvious changes, but I cannot run the script locally. Running .\build.ps1 --target=Test fails with error CS0234: The type or namespace name 'Git' does not exist in the namespace 'Cake' (are you missing an assembly reference?). I've tried to make a PR just to see whether it runs on the CI
Full trace
.\build.ps1 --target=Test
Tool 'cake.tool' (version '5.0.0') was restored. Available commands: dotnet-cake
Tool 'gitversion.tool' (version '6.4.0') was restored. Available commands: dotnet-gitversion
Tool 'gitreleasemanager.tool' (version '0.20.0') was restored. Available commands: dotnet-gitreleasemanager
Restore was successful.
Tool GitVersion.Tool is already installed, with required version.
Tool GitReleaseManager.Tool is already installed, with required version.
.NETStandard,Version=v2.0
error CS0234: The type or namespace name 'Git' does not exist in the namespace 'Cake' (are you missing an assembly reference?)
Error: Error(s) occurred when compiling build script:
error CS0234: The type or namespace name 'Git' does not exist in the namespace 'Cake' (are you missing an assembly reference?)
@mikkelbu It appears to have built OK in CI. This may be something I should check in the recipe. The build wrappers only restore the tools but they have to be installed first. I'll look into this separately since the recipe is a separate repo.
@CharliePoole One detail from our reproduction that may help: Cake.Git is loaded through the recipe’s #addin directive, separately from the manifest tools. dotnet tool restore itself installs the listed local tools; prior installation isn’t required. See the Microsoft documentation and Cake addin documentation.
Our fresh review checkout restored Cake.Git 5.0.1 and compiled past the missing-namespace error. It subsequently stopped at a separate GitVersion history error. That doesn’t establish what failed on Mikkel’s machine, but suggests checking addin restoration/reference loading separately from tool-manifest restoration.
Prepared by Neil Colvin with assistance from OpenAI Codex.
Please check what's in version3 head to make sure you have everything you want for the adapter.
As far as I can see we have #1927 which would otherwise be a blocker and if #1967 (Issue #1966) comes in too that would be great.
I am not so concerned about this issue (#1954) since it the adapter doesn't seem to need this.
These would then go into the 6.4 adapter, which I want to get out as soon as possible.
@OsirisTerje Both fixes you mention are already in version3: #1927 was fixed by merged #1931, and #1967 merged on 15 November as 84a99af, which is the current version3 head. The branch also contains #1955, restoring NUnit 5 support and its parameterless cooperative-stop entry point.
I checked the ancestry rather than relying only on the issue states. #1970 is based on that same branch head and adds the separate .NET 11 Console agent. This confirms the source fixes are present for the engine version you intend to consume in adapter 6.4; package publication and release sequencing remain to be agreed.
Activity
CharliePoole commentedon Nov 4, 2026
@nunit/engine-team Related to this, I'm retiring the "V4" label but keeping "V3." All open issues not labeled "V3" may be considered as V4 issues in future.
oznetmaster commentedon Nov 4, 2026
@CharliePoole, I support adding a .NET 11 agent if someone can take responsibility for the implementation and release. With V4 still some months away, that would benefit existing V3 Console users.
Our current commitments remain the V4 remote-provider work and adapter integration. I can make our existing .NET 11 fixtures and validation results available to whoever takes this on, but I’m not volunteering for ongoing V3 release maintenance.
Prepared by Neil Colvin with assistance from OpenAI Codex.
CharliePoole commentedon Nov 4, 2026
Thanks @oznetmaster
I don't think any of our work so far on V4 is needed. Agents in v3 are built-in and are simply different runtime targets of a single assembly. Creating one is a matter of adding net11.0 as a target for the agent and its tests and updating the workflow. Making it work is a bit more, but I think pretty trivial. As stated earlier, I suspect net11.0 will issue even more warnings than net10.0, but that's not hard to fix either.
But every new agent I've added triggers more bug reports and I don't doubt this will too. That's the big job for the person who is going to volunteer. :-)
mikkelbu commentedon Nov 5, 2026
I think we should do it - and I can take a look at doing this (possibly this weekend). I cannot guarantee that I'll have time for all the bug reports :-)
oznetmaster commentedon Nov 5, 2026
@mikkelbu, I can share our existing .NET 11 fixtures and expected outcomes for validation of the V3 built-in agent.
One separate V3 compatibility fix to keep in view is draft #1955: it adapts the modern driver to NUnit 5’s parameterless StopRun API while preserving the older Boolean signature. It does not add a .NET 11 agent or depend on that work, but could be considered for the next V3 release once reviewed.
Our main implementation commitments remain V4 remote execution and adapter integration.
Prepared by Neil Colvin with assistance from OpenAI Codex.
[-]Will we add a .NET 11.0 agent for V3?[/-][+]Add a .NET 11.0 agent for V3[/+]CharliePoole commentedon Nov 5, 2026
@mikkelbu Thank you! In that case let's do it!
I've changed the title and labeling of this issue and assigned it to you. The version3 branch already contains fixes for #1927 and #1856. I'm re-targetting those to the 3.23 release so they'll appear in the release notes. I'll send a follow-up note to you about how to do the actual release.
@OsirisTerje Please check what's in version3 head to make sure you have everything you want for the adapter.
CharliePoole commentedon Nov 14, 2026
@mikkelbu Were you able to make any progress on this?
mikkelbu commentedon Nov 15, 2026
@CharliePoole Unfortunately not. Last weekend "disappeared" due to illness in my family (nothing major, but just took my time) and since then I've been (too) busy with work. I hope I've some time today, but next is also going to be busy (at least the working days)
CharliePoole commentedon Nov 15, 2026
@mikkelbu I understand completely how that can happen! In fact, it seems like most of the team is completely swamped, so you're not alone.
Anyway, if you can't get to it today, I'd like to put it "up for grabs" again.
mikkelbu commentedon Nov 15, 2026
I only had time to make the obvious changes, but I cannot run the script locally. Running
.\build.ps1 --target=Testfails witherror CS0234: The type or namespace name 'Git' does not exist in the namespace 'Cake' (are you missing an assembly reference?). I've tried to make a PR just to see whether it runs on the CIFull trace
CharliePoole commentedon Nov 16, 2026
@mikkelbu It appears to have built OK in CI. This may be something I should check in the recipe. The build wrappers only restore the tools but they have to be installed first. I'll look into this separately since the recipe is a separate repo.
oznetmaster commentedon Nov 16, 2026
@CharliePoole One detail from our reproduction that may help:
Cake.Gitis loaded through the recipe’s#addindirective, separately from the manifest tools.dotnet tool restoreitself installs the listed local tools; prior installation isn’t required. See the Microsoft documentation and Cake addin documentation.Our fresh review checkout restored Cake.Git 5.0.1 and compiled past the missing-namespace error. It subsequently stopped at a separate GitVersion history error. That doesn’t establish what failed on Mikkel’s machine, but suggests checking addin restoration/reference loading separately from tool-manifest restoration.
Prepared by Neil Colvin with assistance from OpenAI Codex.
OsirisTerje commentedon Nov 16, 2026
As far as I can see we have #1927 which would otherwise be a blocker and if #1967 (Issue #1966) comes in too that would be great.
I am not so concerned about this issue (#1954) since it the adapter doesn't seem to need this.
These would then go into the 6.4 adapter, which I want to get out as soon as possible.
oznetmaster commentedon Nov 16, 2026
@OsirisTerje Both fixes you mention are already in version3: #1927 was fixed by merged #1931, and #1967 merged on 15 November as 84a99af, which is the current version3 head. The branch also contains #1955, restoring NUnit 5 support and its parameterless cooperative-stop entry point.
I checked the ancestry rather than relying only on the issue states. #1970 is based on that same branch head and adds the separate .NET 11 Console agent. This confirms the source fixes are present for the engine version you intend to consume in adapter 6.4; package publication and release sequencing remain to be agreed.
Prepared jointly by Neil Colvin and OpenAI Codex.