16:46 <SeanTAllen> jemc: i discussed a way with Sylvan to make FFI using code more testable. Idea would be for the normal FFI code for something like TCPConnection be 1 or more primitives that do "the normal" FFI calls

16:46 <SeanTAllen> and make that a type parameter (the default) to TCPConnection

16:47 <SeanTAllen> then we could, for testing, insert various other handlers for those calls so we can control data returned from received or whatever else to be able to better test those scenarios

16:47 <SeanTAllen> could also allow for injecting failures etc

16:47 <SeanTAllen> when i have some time i'll play around more with the idea

16:59 <jemc> yeah, that's a good idea

16:59 <jemc> wrapping FFI calls in a primitive also just seems like good organization practice

16:59 <jemc> to keep the FFI code more isolated

17:01 <SeanTAllen> agreed

17:03 <TwoNotes> How about putting all those primitives in a single .pony file within a package, so all the FFI stuff is in one place?

17:05 <SeanTAllen> i dont like that idea at all

17:05 <SeanTAllen> i shouldnt need to pull in primitives for file FFI to do socket work

17:06 <SeanTAllen> if there is FFI that requires regex, I dont want to have to compile in PCRE in order to do sockets etc

17:06 <TwoNotes> I meant within a single package

17:06 <SeanTAllen> i would expect file related ones to be with the other file code, net etc

20:26 <staticassert> Sorry to miss the sync meeting, retrospective is scheduled right when it happens. I'll be sure to listen in to the recording.