Compiler SIGSEGV with generic term rewriting template

[ghost]

This code:

template rewriteMinusOne{(a = a + 1)|(a += 1)}[T](a: T): untyped = 
  a -= T(1)

var a = 3

a = a + 1
echo a
a += 1
echo a

crashes the compiler. Traceback:

nim.nim(121)             nim
nim.nim(77)              handleCmdLine
main.nim(163)            mainCommand
main.nim(74)             commandCompileToC
modules.nim(240)         compileProject
modules.nim(180)         compileModule
passes.nim(217)          processModule
passes.nim(135)          processTopLevelStmt
sem.nim(550)             myProcess
sem.nim(531)             semStmtAndGenerateGenerics
hlo.nim(103)             hloStmt
hlo.nim(75)              hlo
hlo.nim(56)              applyPatterns
hlo.nim(27)              evalPattern
semexprs.nim(18)         semTemplateExpr
evaltempl.nim(157)       evalTemplate
evaltempl.nim(59)        evalTemplateAux
evaltempl.nim(59)        evalTemplateAux
evaltempl.nim(43)        evalTemplateAux
system.nim(2737)         sysFatal
Error: unhandled exception: index out of bounds [IndexError]

I know that this snippet is stupid and doesn't make any sense. However it works with a -= 1 instead of a -= T(1)

[krux02]

The orror occurs in compiler/evaltempl.nim(44): index 1 not in 0 .. 0. @Araq, to fix it, I request some documentation for evalTemplateAux. What should this function do precisely? What are all the arguments? Also a better documentation for the fields in TemplCtx would be nice.

[ghost]

With latest devel it produces an error instead:

/home/dian/Projects/nimforum/kk.nim(6, 3) template/generic instantiation of `rewriteMinusOne` from here
/home/dian/Projects/nimforum/kk.nim(1, 54) Error: undeclared identifier: 'T'

[narimiran]

With latest devel it produces an error instead:

July 2020 update:

Error: unhandled exception: index 1 not in 0 .. 0 [IndexDefect]

which is the same as reported originally.

[bung87]

in 1.6.8 Error: unhandled exception: index 1 not in 0 .. 0 [IndexDefect]

think this can be close ?

[ringabout]

think this can be close ?

No, it is still a crash.

[bung87]

what's the expected, I dont think localError == crash