No error for extern objects with constructor name matching a different type

[jafingerhut]

Happened across this while puzzling through what TYPE meant in the grammar rule for constructor method declarations. I don't think this is a high priority issue, but wanted to file one in case someone else thought it was worth a change in the compiler. See comments inside myExtern1 in the attached program.
wrong-constructor-name.p4.txt

[jafingerhut]

Additional comment/question -- the latest P4_16 draft spec says: "Extern declarations can only appear in architecture model libraries specific to a target." p4c doesn't seem to enforce this. Does that statement look correct?

If architecture model libraries are in #include files, it doesn't seem like there is even a very straightforward way for the compiler to enforce that restriction?

[mihaibudiu]

Indeed, the compiler cannot enforce this.
We may want to remove this phrase from the spec; Netronome has this use for extern declarations where they indicate user-supplied functions in C which are linked to the P4 program. In this case users can write new extern declarations. Maybe we should add a special annotations to indicate to the back-end that it does not need to understand them.

[jafingerhut]

Confirmed that the test case now gives an error message for the thing that looks like a constructor, but has a different type name than the extern object it is within. Thanks, Mihai.