search

cpsc411 / cpsc411-book

The WIP book for CPSC 411
Other
58 stars 7 forks source link

Elaboration on info field semantics #110

Open wilbowma opened 1 year ago

wilbowma commented 1 year ago

https://piazza.com/class/lcquw2vc2cs73i/post/195 I'll consider this a bug. Any existing info fields should have been preserved, since the source language contains an info field, and the info field is only a partial spec (technically, it's meant to be understood as a row-polymorphic record). But the reference compiler discards the additional entries.

However, it's reasonable for uncover-locals to throw away an existing locals field. The specification for uncover-locals tells us that it introduces that field, officially, to the compiler, and what the field contains. uncover-locals must be responsible for that field, and it cannot trust any values that were in there before, or a bug caused by a bad locals field would risk being blamed on uncover-locals unjustly. The same reason applies to any pass that introduces an info field.

wilbowma commented 1 year ago

The bug is fixed, but I need to elaborate on these semantics in the book