Better diagnostics for coreclr initialization failures

[janvorli]

When coreclr_initialize fails, the returned exit code is most often the generic 0x80004005 (E_FAIL). To diagnose such failures, runninga repro under a debugger is the only viable option now.
We should add some extra logging so that a host can report more details on such a failure that could pinpoint the error source quickly, without a need to execute a repro under a debugger.

[richlander]

That sounds good! I have sometimes use COREHOST_TRACE to get information on what went wrong. Would you expect this extra information to be included in that tracing? If not, where?

[AaronRobinsonMSFT]

/cc @elinor-fung

[janvorli]

@richlander I would make this extra info part of the message that currently reports the coreclr initialization failure. We currently print this into console:

Failed to initialize CoreCLR, HRESULT: 0x80004005

I was thinking that it would be sufficient to add some extra number identifying the source of the failure. Alternatively, we could possibly generate a special range of HRESULT codes that we would use instead of the E_FAIL.

[richlander]

SGTM ... look forward to seeing a demo. I love error messages.

[janvorli]

Fixed by #78790