-
Notifications
You must be signed in to change notification settings - Fork 10k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrade to more recent .NET SDK #21017
Conversation
I believe the new SDK brings along a number of breaking changes. New errors and warnings include
The final error implies native assets aren't being copied where they're expected. So, @ViktorHofer @dsplaisted @rainersigwald @rrelyea what's changed and how do we need to react in this repo? |
|
FYI we don't do much more than the following in the projects showing NU5129 warnings.
Are you saying the fix is to change our |
what build of sdk were you using before? what version are you upgrading to? where you see us treating netcoreapp5.0 as net5.0, and you are running into problems... until then, you may be able to live with ignoring the warning (via nowarn)? |
@rrelyea this PR makes the following change - "version": "5.0.100-preview.2.20120.3"
+ "version": "5.0.100-preview.4.20202.8" |
That won't solve the error shown above in #21017 (comment). |
by error, you mean the exception? we generally tried to leave netcoreapp5.0 working in parallel with net5.0. so you need to figure out if you can move forward, or if you need to stay back. when does Microsoft.NetCore.App get fixed and by who? |
@rrelyea why would
@ericstj @ViktorHofer I don't see any mention of If that's not coming soon, we can't change our default TFM until we get VS16.6 on the agents or I resolve the problems seen in #20748. The short-term workaround would be to update |
a416fb0
to
80e92e8
Compare
We will switch to net5.0 on May 4th: dotnet/runtime#35202 |
My added logging is dumping into |
There are a couple of things going on here:
|
@dougbu The change to generate the However, the definition of
You could use |
We also need to edit the .deps file. adding information about a native assembly. That's why our command executes after Since simplifying |
But,
sounds better to me. |
… because that restores the previous values and customers may also be depending on them. |
@dougbu please update to |
🆙📅 to grab latest from 'master' branch and fix the |
Same or similar issue repros locally. Does a new SDK bring a new hostfxr? I am attached to the hung process right now and it is stuck inside of native code in libcoreclr.dylib, and I noticed that there are some recent changes to that section of code in the Runtime. So I'd like to know if the host can be updated by a SDK update since this PR isn't bringing in a new Runtime. |
Oh, the shared runtime did update, I didn't notice that. So, this might be a regression in hostfxr. |
Huh, I have no idea anymore, I get the hanging tests on master branch too on my Mac. |
@BrennanConroy did you delete the .dotnet/ folder before trying again in 'master'? Some parts of the build seem to roll forward when multiple SDKs are available on dev machines. This probably means |
./clean.sh deletes .dotnet/ right? |
Nope, |
|
Oh wait, I did a manual |
And, did you burn any leftover |
Yep (except for a zombie dotnet that just wont go away, but it's a zombie so it doesn't affect anything) |
@BrennanConroy zombies 🧟 are not to be trifled with.
More seriously, what is your Mac doing exactly when the timeouts occur❔ And, because we're not seeing similar problems in the rolling builds of our 'master' branch, I wonder about scorched Earth approaches like clearing your NuGet caches and rebooting. |
/fyi latest validation build (https://dev.azure.com/dnceng/public/_build/results?buildId=630538) showed some improvement. But,
Bottom line, the remaining test failures are specific to quarantined tests but they are also consistent enough to imply we may still be dealing with product issues such as the potential |
I'm doubting this is the case anymore |
Any new and improved ideas❔ |
Nope, my machine is unreliable since I can't even complete the tests on the master branch |
@dougbu what makes you think the error has to do anything with the certificate? That error merely implies that the app is failing to start at all. |
@javiercn the last server message before the client gives up is
The server doesn't get around to emitting its usual "ready" messages. In addition, hasn't the notarization logic in the CLI and SDK changed since 5.0 preview 2? |
Yeah, sorry. I lost track of this PR and couldn't answer. I already looked into it and there was an issue with the cert loading code. I put out a fix and improved diagnostics and I'll merge it soon. |
Yes, that's why we are seeing this issue I think. The server is not picking up the certificate we generate I believe. |
Your change is only for the Templates code, however some Kestrel tests are hanging with a similar issue it looks like. They are trying to access the dev cert and never completing. Were there recent changes to how the cert code works? Do we need to do some manual cert injection for the tests in Kestrel that use https? |
/azp run |
Azure Pipelines successfully started running 2 pipeline(s). |
Yay, Mac passed 😄 |
🎉 🎉 🎉 |
No description provided.