[WIP] SILGen: Don't inject "willThrow" hooks before rethrow propagation branches.

[jckarter]

The hook is intended to be used by debuggers to catch the point a throw happened in user source. It's unnecessary and undesirable to hook in places where an already-thrown error is just being implicitly propagated.

[jckarter]

@swift-ci Please test

[jckarter]

@adrian-prantl @jasonmolenda Do you recall whether it was intentional that we emitted willThrow hooks before implicit propagation like this? IIRC we only intended to call willThrow before an explicit throw in source code.

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 9207930fc1d74130d7e7a1938e200cdf77d5eeb1

[adrian-prantl]

I'm afraid I don't know anything about this. Was the commit history helpful at all?

[swift-ci]

Build failed
Swift Test Linux Platform
Git Sha - 9207930fc1d74130d7e7a1938e200cdf77d5eeb1

[jckarter]

It looks like the willThrow call was put there in @rjmccall's original commit 58b5e1d and @atrick's 482b264 without any specific commentary. I spoke with @rjmccall, and his recollection matched mine.

[jckarter]

@swift-ci Please test

[swift-ci]

Build failed
Swift Test Linux Platform
Git Sha - 9207930fc1d74130d7e7a1938e200cdf77d5eeb1

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 9207930fc1d74130d7e7a1938e200cdf77d5eeb1

[rjmccall]

Yes, I think this is reasonable.

You could make an argument that we should call willThrow before starting the propagation of an error received from ObjC, just so that such throws do always get trapped. However, that would still sweep up a fair amount of propagation-rethrows and expose the details of exactly how we performed a call. I think the "it's only triggered by Swift throw statements" line is something that users will understand.

[jckarter]

ObjC-to-Swift propagation has its own code paths, so we could conceivably still emit the hook when propagating in that situation.

[jckarter]

I discussed this a bit with @jimingham on swift-lldb-dev. We should keep the willThrow call after ObjC calls that bridge to throw.

[CodaFi]

@jckarter Do you think this would make for a good starter bug in IRGen?

[jckarter]

Yeah, that's a good idea!

[CodaFi]

Filed https://bugs.swift.org/browse/SR-11803. Please let me know if I'm missing any details in there.

Going to close this out.