There appears to be a regression in cgo, where when a C function is called with not enough arguments, it causes cgo to panic instead of printing a meaningful error:

package main

// #include <math.h>
// #cgo LDFLAGS: -lm
import "C"

func main() {
    _ = C.sin()
}

In go 1.4 go build prints out:
./foo.go:8: not enough arguments in call to _Cfunc_sin

At tip, the output is a panic during the execution of cgo:

panic: runtime error: index out of range

goroutine 1 [running]:
main.(*Package).rewriteCall(0xc82009e0f0, 0xc820082120, 0xc820076b00, 0xc8200a0180)
    /Users/cookieo9/local/go/src/cmd/cgo/gcc.go:606 +0xc58
main.(*Package).rewriteCalls(0xc82009e0f0, 0xc820082120)
    /Users/cookieo9/local/go/src/cmd/cgo/gcc.go:593 +0x1f8
main.(*Package).Translate(0xc82009e0f0, 0xc820082120)
    /Users/cookieo9/local/go/src/cmd/cgo/gcc.go:170 +0x197
main.main()
    /Users/cookieo9/local/go/src/cmd/cgo/main.go:275 +0x1055

Going the other way (i.e.: having too many arguments) produces the same error as go 1.4:
./foo.go:8: too many arguments in call to _Cfunc_sin

The only function I tested which doesn't panic, and prints the correct error is malloc, although I believe that's a result of special handling for malloc. Other functions I tested included "strcpy", "strlen", and "calloc" (from https://groups.google.com/forum/#!topic/golang-dev/TquSAX-w6NA).

I have tested this at tip on darwin-amd64 (El Capitain), and linux-arm (Raspberry Pi 2)