Repository navigation
Unintended dead code in binary schema generation #295
Description
Activity
The different compilers frustratingly each find totally disjoint sets of unreachable blocks, attesting to just how unreliable these warnings are, but I raised this one from msvc to your attention because it definitely seemed both valid and a genuine bug. Looks a bit like a merge that went awry.
clang found a bunch of other unreachable blocks which seem genuinely unreachable but don't appear malign: the
gen_panicmacro in src/compiler/codegen_c.h which internallyasserts and thenexits, so any code after it in a basic block is unreachable.Incidentally, gcc accepts the
-Wunreachable-codeflag but it has no effect on the compiler's behaviour, only being accepted for reasons of backwards compatibility according to their doc.There is a some standard expression or attribute to indicate a function that does not (always) return.
I don't recall, but if added, portable library would need to manage it for when compilers don't know about it.
There seems to be unreachable code in src/compiler/codegen_schema.c.
msvc looks at this block between the two
breaks and raises a warning. From a glance at the code, its analysis seems valid. I know this isn't strictly related to this PR, but I thought I would bring it to your attention as the code definitely looks strange to me.Originally posted by @amcn in #262 (comment)
EDIT: the dead code prevent adding a binary schema entry for string array struct fields. This is a FlatCC only feature that should have been mapped to a ubyte array in the schema, but the code was never completed for that (it will be when this issues is fixed).