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
{{ message }}
Repository navigation
The productive version of Microsoft.Testing.Extensions.HangDump must not reference to a preview version of Microsoft.Diagnostics.NETCore.Client #11981
The package Microsoft.Testing.Extensions.HangDump in version 2.5.1 references to the package Microsoft.Diagnostics.NETCore.Client in version 0.2.607501, which is assemed to be a preview or alpha version.
A productive package like HangDump should reference to productive packages only.
Version used
Microsoft.Testing.Extensions.HangDump 2.5.1
Steps To Reproduce
N/A
Expected behavior
All references of Microsoft.Testing.Extensions.HangDump have the same release state as the package itself.
Actual behavior
Microsoft.Testing.Extensions.HangDump seems to be productive, but refers to a preview version.
Additional context
We have clear strategy to consume packages from NuGet.org. Packages are checked about the current development state. Packages should be consistent and under maintenance.
While Microsoft.Testing.Extensions.HangDump is assumed to be well maintained, the reference to Microsoft.Diagnostics.NETCore.Client 0.2.607501 breaks the chain of trust.
Sorry for the delay, I was OOF sick last week. I am continuing the discussion with internal teams. So far, the answer seems to be that this is indeed not considered as stable but "should not be a problem".
Do you have some strong requirement on your project/org/company or is it more of a "on principle" discussion?
We (my company) have a package approval process. Every package, that is intended to be included in our internal (approved) package feed shall fulfill some requirements, like maintained and (quite) stable.
For this package we need to define an exception. So, it is more than an "on principle" discussion.
A package denoted as stable (due to version >= 1.0) should not refer to a package denoted as unstable. If it does, it cannot be really stable.
If Microsoft.Diagnostics.NETCore.Client is working as expected, at least for Microsoft.Testing.Extensions.HangDump, is can be defined as stable, isn't it?
If Microsoft.Diagnostics.NETCore.Client is working as expected, at least for Microsoft.Testing.Extensions.HangDump, is can be defined as stable, isn't it?
In theory yes :) Reality in big corps with many different teams is a little more complex but I'll push on the teams.
We have been able to circle back on this. The package is currently maintained but not yet shipped as v1+ because they consider the API still unstable (although there was no break in a long time). The diagnostic team will work on stabilizing the API so they can ship v1 but that might take some time.
In the meantime, the package is correctly following semantic versioning:
it's not a preview package (it's maintained)
it's not yet v1 because API is unstable (but maintained)
maintenance includes security/vulnerability fixes
If some breaks in the APIs happens, this is transparent to users of MTP because this is a deps of our HangDump extension so we would have to fix things on our side.
The key distinction here is (un)stable vs (un)maintained: package Microsoft.Diagnostics.NETCore.Client is still unstable but it is fully supported/maintained/secured.
Activity
Evangelink commentedon Oct 16, 2026
Hello!
Thanks for the report, I'll reach out to internal team but I am not seeing any mention that this is a preview version.
lg2de commentedon Oct 16, 2026
If so, then please give a version 1.x.
Evangelink commentedon Oct 26, 2026
Sorry for the delay, I was OOF sick last week. I am continuing the discussion with internal teams. So far, the answer seems to be that this is indeed not considered as stable but "should not be a problem".
Do you have some strong requirement on your project/org/company or is it more of a "on principle" discussion?
lg2de commentedon Oct 26, 2026
We (my company) have a package approval process. Every package, that is intended to be included in our internal (approved) package feed shall fulfill some requirements, like maintained and (quite) stable.
For this package we need to define an exception. So, it is more than an "on principle" discussion.
A package denoted as stable (due to version >= 1.0) should not refer to a package denoted as unstable. If it does, it cannot be really stable.
If Microsoft.Diagnostics.NETCore.Client is working as expected, at least for Microsoft.Testing.Extensions.HangDump, is can be defined as stable, isn't it?
Evangelink commentedon Oct 27, 2026
In theory yes :) Reality in big corps with many different teams is a little more complex but I'll push on the teams.
Evangelink commentedon Oct 29, 2026
We have been able to circle back on this. The package is currently maintained but not yet shipped as v1+ because they consider the API still unstable (although there was no break in a long time). The diagnostic team will work on stabilizing the API so they can ship v1 but that might take some time.
In the meantime, the package is correctly following semantic versioning:
If some breaks in the APIs happens, this is transparent to users of MTP because this is a deps of our HangDump extension so we would have to fix things on our side.
The key distinction here is (un)stable vs (un)maintained: package
Microsoft.Diagnostics.NETCore.Clientis still unstable but it is fully supported/maintained/secured.