`nim check` doesn't report some compiler errors (eg: static array bound check)

[timotheecour]

#[
KEY nim check

`nim check` doesn't report some compiler errors

nim check $timn_D/bugs/all/t0053.nim
passes (BUG!)

nim c -o:/tmp/z01 $timn_D/bugs/all/t0053.nim
t0053.nim(10, 6) Error: index out of bounds

## note: with debug info,
$nim_dev_debug_X c -o:/tmp/z01 $timn_D/bugs/all/t0053.nim
gives:
Traceback (most recent call last)
/Users/timothee/git_clone/nim/Nim/compiler/nim.nim(109) nim
/Users/timothee/git_clone/nim/Nim/compiler/nim.nim(73) handleCmdLine
/Users/timothee/git_clone/nim/Nim/compiler/cmdlinehelper.nim(91) loadConfigsAndRunMainCommand
/Users/timothee/git_clone/nim/Nim/compiler/main.nim(168) mainCommand
/Users/timothee/git_clone/nim/Nim/compiler/main.nim(78) commandCompileToC
/Users/timothee/git_clone/nim/Nim/compiler/modules.nim(134) compileProject
/Users/timothee/git_clone/nim/Nim/compiler/modules.nim(79) compileModule
/Users/timothee/git_clone/nim/Nim/compiler/passes.nim(194) processModule
/Users/timothee/git_clone/nim/Nim/compiler/passes.nim(86) processTopLevelStmt
/Users/timothee/git_clone/nim/Nim/compiler/cgen.nim(1487) myProcess
/Users/timothee/git_clone/nim/Nim/compiler/ccgstmts.nim(1186) genStmts
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2408) expr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2201) genStmtList
/Users/timothee/git_clone/nim/Nim/compiler/ccgstmts.nim(1186) genStmts
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2370) expr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2039) genMagicExpr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(1020) genEcho
/Users/timothee/git_clone/nim/Nim/compiler/cgen.nim(504) initLocExpr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2390) expr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2166) genArrayConstr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2376) expr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(1987) genMagicExpr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(1538) genDollar
/Users/timothee/git_clone/nim/Nim/compiler/cgen.nim(504) initLocExpr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(2402) expr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(951) genBracketExpr
/Users/timothee/git_clone/nim/Nim/compiler/ccgexprs.nim(878) genArrayElem
/Users/timothee/git_clone/nim/Nim/compiler/msgs.nim(522) localError
/Users/timothee/git_clone/nim/Nim/compiler/msgs.nim(504) liMessage
/Users/timothee/git_clone/nim/Nim/compiler/msgs.nim(332) handleError
/Users/timothee/git_clone/nim/Nim/compiler/msgs.nim(317) quit
]#

let a = [1,2]
echo a[3]

why this is annoying

• nim check is very useful to report multiple errors in a single test case (eg in testament); without this, we can't rely on nim check and have to use 1 test case per unique error (and 1 separate compilation, which is slow)
• nim check is often used in editor integrations, eg NimLime uses that to report any errors

[Araq]

That's what you get when the backend is allowed to report errors. You would have written it the same way because of "DRY".

[timotheecour]

@araq follow up from https://github.com/nim-lang/Nim/pull/9916#issuecomment-445839902 :

IMO system.compiles's spec should be "type supports operation" as well as "is importable module", nothing more

so compiles("[0,1][10]") returning true is intended behavior ? docs could be a bit clarified regarding that aspect (from user POV, it's not clear that it's a backend vs semantic error)

in any case, I agree with your point thatcompiles(s) shouldn't call into backend ; and also should be kept as efficient as possible, since otherwise it could slow down compilation for some programming patterns (eg toSeq)

[Araq]

compiles("[0,1][10]") is an edge case that could return either. But in theory array indexing is a "supported operation" so true fits the bill.