Mirror of forum.nim-lang.org

373 :: Forwarding varargs to C

• simenss (orginal) [2014-02-18T22:42:19+01:00] view original

  I just discovered using macros with varargs[expr] which strikes me as extremely powerful and, for some of my use cases, potentially extremely efficient compared to my current C implementations. But right now I need to wrap a C function with some additional functionality. As far as I can tell, I can't use a regular proc in any way, since my arguments are of different types, so I would have to use a macro. Is there any way I can forward a varargs-list to a printf-style C function via a macro (or template)?

---

• gradha (orginal) [2014-02-19T00:23:43+01:00] view original

  You could try using the varargs pragma which provides a printf example.

---

• simenss (orginal) [2014-02-19T00:50:45+01:00] view original

  True, I am allready able to call my varargs-function from Nimrod. It is declared like this:

  proc write*(bq : PBlockQueue, fmtstr : cstring) : void {.cdecl,
    importc: "bqueue_write", varargs.}
  Run

  The problem seems to be that the varargs isn't actually treated as anything in the Nimrod end of things. When I try to wrap this function with some extra functionality on the Nimrod side, like this:

  template write*(d : var TDisplay, node : int, args : varargs[expr]) =
    d.address(node)
    d.pipe.write(args)
  Run

  I get told off by the compiler:

  display.nim(51, 9) Error: type mismatch: got (PBlockQueue, varargs[expr])
  but expected one of:
  bqueue.write(bq: PBlockQueue, fmtstr: cstring)
  Run

  Which is the function I want to call, but I see no way to match it with my varargs since they are not formally part of the signature, it seems.

---

• LeuGim (orginal) [2014-02-19T02:10:55+01:00] view original

  template write*(d : var TDisplay, node : int, fmtstr: cstring, args : varargs[expr]) =
    d.address(node)
    d.pipe.write(fmtstr, args)
  Run

  This works for me. Just 3rd argument missed.

---

• simenss (orginal) [2014-02-21T08:05:55+01:00] view original

  Thanks for the tip. Makes sense, I should have thought of that myself. Thought varags would "file into" the arguments since the first parameter of the varargs would me fmtstr. Anyway, must have tripped on a compiler bug, because now I get:

  Error: internal error: getTypeDescAux(tyExpr)
  No stack traceback available
  Run

  Odd. It seems the compiler is unable to find a type name for my 'exprs': https://github.com/Araq/Nimrod/blob/2fc928688eab6ad389ae1fce87230f1ecba69106/compiler/ccgtypes.nim#L613

---

• simenss (orginal) [2014-02-21T08:45:28+01:00] view original

  Tried again with the latest version from github, so now I have a traceback:

  cgen.nim(1289)           myProcess
  ccgstmts.nim(995)        genStmts
  ccgexprs.nim(1902)       expr
  ccgstmts.nim(995)        genStmts
  ccgexprs.nim(1966)       expr
  cgen.nim(880)            genProc
  cgen.nim(850)            genProcNoForward
  cgen.nim(784)            genProcAux
  ccgstmts.nim(995)        genStmts
  ccgexprs.nim(1902)       expr
  ccgstmts.nim(995)        genStmts
  ccgexprs.nim(1902)       expr
  ccgstmts.nim(995)        genStmts
  ccgexprs.nim(1902)       expr
  ccgstmts.nim(995)        genStmts
  ccgexprs.nim(1858)       expr
  ccgcalls.nim(294)        genCall
  ccgcalls.nim(138)        genPrefixCall
  ccgcalls.nim(118)        genArgNoParam
  cgen.nim(586)            initLocExpr
  ccgexprs.nim(1876)       expr
  ccgexprs.nim(1713)       genArrayConstr
  cgen.nim(406)            getTemp
  ccgtypes.nim(620)        getTypeDesc
  ccgtypes.nim(585)        getTypeDescAux
  ccgtypes.nim(613)        getTypeDescAux
  msgs.nim(826)            internalError
  msgs.nim(764)            rawMessage
  msgs.nim(761)            rawMessage
  msgs.nim(714)            handleError
  Run

---

• LeuGim (orginal) [2014-02-21T14:06:07+01:00] view original

  I don't know how to reproduce this error, I haven't seen your code. The minimal code that I recreated is at https://github.com/Leu-Gim/simenss, it works (my compiler is about 25 days old, from devel).

---