Support special casing with Fabric endpoints

[ErikEJ]

Support the special casing for Fabric endpoints

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 72.88%. Comparing base (b8e0c9f) to head (4c73ef8).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3084      +/-   ##
==========================================
+ Coverage   72.70%   72.88%   +0.18%     
==========================================
  Files         283      283              
  Lines       58975    58978       +3     
==========================================
+ Hits        42875    42984     +109     
+ Misses      16100    15994     -106     

Flag	Coverage Δ
addons	92.58% <ø> (ø)
netcore	75.58% <100.00%> (+0.07%)	⬆️
netfx	71.33% <100.00%> (+0.19%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[edwardneal]

Only one comment on the code style, but the checks themselves are correct.

I think there are a handful of feature differences between Azure SQL and Fabric SQL which merit a test suite and documentation updates though. At a glance, the driver would be impacted by the lack of support for Always Encrypted and the BULK INSERT statement.

[edwardneal]

Perhaps this should be a resource string?

[ErikEJ]

Yeah, wonder why the other ones are resource strings, I guess they are not localized ?

[ErikEJ]

@edwardneal There are a number of feature differences between the many engine variations this driver supports or happen to work with - is it really the responsibility of the driver to document this, or is the page you are linking to not enough?

[edwardneal]

I think the page I linked to would cover most cases, I'm thinking primarily about a once-over to make sure that the SqlClient documentation doesn't contradict it and is clear enough for people to understand what functionality the feature depends upon. The example I had in mind was SqlBulkCopy - it might not be obvious that the lack of BULK INSERT support blocks this feature. I think that's probably the root cause of #2995.

[ErikEJ]

#2995 refers to Fabric warehouse, which is something completely different from Fabric SQL Database

[edwardneal]

Sorry, slightly delayed here!

You're right, but the root cause is similar: some of SqlClient's functionality can't be supported on a particular engine variation, because that functionality relies upon a feature the engine doesn't support (in this case, SqlBulkCopy relies upon the BULK INSERT statement, which neither Fabric Warehouse nor Fabric SQL Database support.)

My broader point is that we should make sure that the SqlClient documentation for functionality which depends on BULK INSERT or Always Encrypted doesn't claim to be compatible across all engine variations, and that specifically for SqlBulkCopy we should make it clear that the feature depends upon server-side BULK INSERT support (since that's hard to discover in the documentation, and without that insight it's hard to know what won't support the feature.) If you want to keep this PR scoped specifically to the endpoint detection, I'm happy enough to check that documentation in a separate PR.

[ErikEJ]

Sounds to me like a separate issue and PR would be suitable to prevent scope creep.