Spmi-diff is showing diffs that are not related to the PR

[kunalspathak]

I believe there is something wrong with spmi-diff pipeline. Even for PRs that shouldn't have code diffs is showing diffs. 

• Pipelines - Run 20220819.5 (azure.com) shows x64 diffs for my change in Use the register of CAST for contained index operator by kunalspathak · Pull Request #74275 · dotnet/runtime (github.com) that was just done under TARGET_ARM64.
• Pipelines - Run 20220819.2 (azure.com) shows diff for a change in that is mostly renaming macros (and hoping that nothing missed out that would trigger the diff)
• And FINALLY, the biggest one Pipelines - Run 20220819.12 (azure.com) shows -16436 code size improvements from Egor Bogatov Improve JitDisasm for top-level statements, expose DumpJittedMethods in Release by EgorBo · Pull Request #74090 · dotnet/runtime (github.com).

I looked into the spmi-diff for #74090 and the diffs looks unrelated. The following diff is coming from #74110, but is reported in #74090.

Likewise, below diff is unrelated to #74090 as well as #74110.

[kunalspathak]

@dotnet/jit-contrib

[dotnet-issue-labeler]

I couldn't figure out the best area label to add to this issue. If you have write-permissions please help me learn by adding exactly one area label.

[AndyAyersMS]

Interesting. The second diff looks like RBO. Maybe the baseline jit is not getting updated or something?

[kunalspathak]

I did check the GUIDs and they look accurate. For example, #74090 the baseline is supposed to be 1b9551b8-21f4-4233-9c90-f3eabd6a322b which is the case as seen in https://helixre107v0xdeko0k025g8.blob.core.windows.net/dotnet-runtime-refs-pull-74090-merge-f0df5552204b43709e/win-x64/1/console.56c5c11e.log?helixlogtype=result.

Even the superpmi-collect seems to be uploading to the right guid. This one is from last successful run.

[21:50:55] Using JIT/EE Version from jiteeversionguid.h: 1b9551b8-21f4-4233-9c90-f3eabd6a322b
[21:50:55] SuperPMI upload

[kunalspathak]

Seems like those are GUIDs for .mch. I don't see logging for baseline jit. May be we should add it.

[AndyAyersMS]

Locally (making changes on top of 9cd1a32) I don't see any spurious diffs.

[kunalspathak]

By the way, from past few days, I am seeing this error on AzDo, not sure what it meant:

[kunalspathak]

Locally (making changes on top of 9cd1a32) I don't see any spurious diffs.

And for that even superpmi diff is clean. https://dev.azure.com/dnceng/public/_build/results?buildId=1950687&view=ms.vss-build-web.run-extensions-tab

[AndyAyersMS]

Look in the "setup" leg for the jit download from the cloud -- eg from https://dev.azure.com/dnceng/public/_build/results?buildId=1954972&view=logs&jobId=011e1ec8-6569-5e69-4f06-baf193d1351e&j=011e1ec8-6569-5e69-4f06-baf193d1351e&t=bf6cf4cf-6432-59cf-d384-6b3bcf32ede2

We should surface this warning more visibly

[AndyAyersMS]

So evidently, we're missing some baseline jits.

[kunalspathak]

Look in the "setup" leg for the jit download from the cloud -- eg from https://dev.azure.com/dnceng/public/_build/results?buildId=1954972&view=logs&jobId=011e1ec8-6569-5e69-4f06-baf193d1351e&j=011e1ec8-6569-5e69-4f06-baf193d1351e&t=bf6cf4cf-6432-59cf-d384-6b3bcf32ede2

We should surface this warning more visibly

Yep, I was trying to find where that is logged. Thanks for finding it.

[kunalspathak]

I see that they are getting uploaded correctly in the rolling jit build job.

[kunalspathak]

he baseline is supposed to be 1b9551b8-21f4-4233-9c90-f3eabd6a322b which is the case as seen

I forgot that for baseline jit, we go with the git_hash.

[BruceForstall]

If I look at the (internal) baseline JIT pipeline https://dev.azure.com/dnceng/internal/_build?definitionId=902, it hasn't built all the JIT changes merged to public 'main'.

[BruceForstall]

We should surface this warning more visibly

It would help if we suppressed some of the download progress status messages.

[ghost]

Tagging subscribers to this area: @dotnet/runtime-infrastructure
See info in area-owners.md if you want to be subscribed.

Issue Details

---

I believe there is something wrong with spmi-diff pipeline. Even for PRs that shouldn't have code diffs is showing diffs. 

• Pipelines - Run 20220819.5 (azure.com) shows x64 diffs for my change in Use the register of CAST for contained index operator by kunalspathak · Pull Request #74275 · dotnet/runtime (github.com) that was just done under TARGET_ARM64.
• Pipelines - Run 20220819.2 (azure.com) shows diff for a change in that is mostly renaming macros (and hoping that nothing missed out that would trigger the diff)
• And FINALLY, the biggest one Pipelines - Run 20220819.12 (azure.com) shows -16436 code size improvements from Egor Bogatov Improve JitDisasm for top-level statements, expose DumpJittedMethods in Release by EgorBo · Pull Request #74090 · dotnet/runtime (github.com).

I looked into the spmi-diff for #74090 and the diffs looks unrelated. The following diff is coming from #74110, but is reported in #74090.

Likewise, below diff is unrelated to #74090 as well as #74110.

Author:	kunalspathak
Assignees:	-
Labels:	area-Infrastructure, untriaged
Milestone:	-

[BruceForstall]

@kunalspathak This has been understood and resolved, so I'm going to close it.