Fix config source gen binding with SslClientAuthenticationOptions

[tarekgh]

Fixes #107003

[tarekgh]

This is a candidate to port to 9.0.

CC @ericstj @eerhardt

[ericstj]

Can you explain all these conditions? I only understand the first, why the others?

[tarekgh]

complexType is not CollectionSpec we have special handling to the collections so we should not change how we do it especially it should work in most cases. look at

runtime/src/libraries/Microsoft.Extensions.Configuration.Binder/gen/Emitter/CoreBindingHelpers.cs

Line 1146 in e4f93db

if (collectionType is not null)

for more details.

((ObjectSpec)complexType).InstantiationStrategy == ObjectInstantiationStrategy.ParameterizedConstructor) is needed because if it is different ObjectInstantiationStrategy value means we can still create and handle the object. look at

runtime/src/libraries/Microsoft.Extensions.Configuration.Binder/gen/Emitter/CoreBindingHelpers.cs

Line 1170 in e4f93db

if (strategy is ObjectInstantiationStrategy.ParameterlessConstructor)

to see that.

[tarekgh]

_typeIndex.GetEffectiveTypeSpec(complexType).IsValueType can be handled too,

runtime/src/libraries/Microsoft.Extensions.Configuration.Binder/gen/Emitter/CoreBindingHelpers.cs

Line 933 in e4f93db

if (effectiveMemberType.IsValueType)

shows what we do with that.

[tarekgh]

I was trying to scope the change to exact failure cases to reduce any risk of breaking anything.

[ericstj]

I see - I was surprised we'd even get this far with a type that has no bindable members and we're trying to construct with a parameterized constructor. I would have expected us to discover that in parsing and use a different spec type (not supported or something) - why do we miss this in parsing?

[tarekgh]

That is because we want use _typeIndex which is created and filled in the emitter. If we want to do the detection earlier, we'll need to do more involving refactoring which will make the change risky for 9.0. Note, the parser fills some data structures which expect to see the type specs supported by _typeIndex later during the emission. so refactoring will not be simple IMO.

[ericstj]

I see - given that it seems like the parser isn't even capable of completely deciding which types will be handled. It can only decide which may be handled and we filter further during emit. That feels like something we'd need to fix to address other issues.

[tarekgh]

/backport to release/9.0

[github-actions]

Started backporting to release/9.0: https://github.com/dotnet/runtime/actions/runs/10818391044