Bug 3746 – Misleading error message "OP has no effect in expression XXX), in void function

On function with void return type, if one will put expression after return: void f() { auto a = 1, b = 3, c = 5; return a/b*c; } will end with compiler error like: aaa.d(4): Error: * has no effect in expression (a / b * c) My function "f" was lot bigger, and took me about 5 minutes to understand what is wrong. Could such bugs can be detected and be something like aaa.d(4): Error: no return statment allowed in void function/method

Well, obviously return *is* allowed from void functions (void f() { return; }). But 'error: return statement with a value in a void function' would be a preferable message than talking about values having no effect in expressions.

(In reply to comment #1) > Well, obviously return *is* allowed from void functions (void f() { return; }). > But 'error: return statement with a value in a void function' would be a > preferable message than talking about values having no effect in expressions. Yes you are right. Error messages should be as precise as possible.

*** Issue 3855 has been marked as a duplicate of this issue. ***

*** This issue has been marked as a duplicate of issue 3630 ***

My bad, this is a request for a better error.

*** Issue 3922 has been marked as a duplicate of this issue. ***

Issue 3922 contains some interesting comments. It seems this enhancement request for a better error message has many dupes!

https://github.com/D-Programming-Language/dmd/pull/121 This adds the error message: testx.d(4): Error: expression with no side effects used with void return Good enough?

(In reply to comment #8) > https://github.com/D-Programming-Language/dmd/pull/121 > > This adds the error message: > testx.d(4): Error: expression with no side effects used with void return > > Good enough? I don't understand. Elsewhere I have a bug report that asks for a new error message if you don't use the result of a pure function. But this bug report just asks for a better error message in the situation of using "return x;" in a void function. So I expected an error message like: foo.d(20): Error: a void function can't return 'int'.

(In reply to comment #9) > > I don't understand. Elsewhere I have a bug report that asks for a new error > message if you don't use the result of a pure function. > But this bug report just asks for a better error message in the situation of > using "return x;" in a void function. So I expected an error message like: > > foo.d(20): Error: a void function can't return 'int'. Returning an int expression from a void function is perfectly legal, so long as the expression has side effects. This is explicitly allowed by the spec. http://www.digitalmars.com/d/2.0/statement.html#ReturnStatement You're talking about two completely separate issues, both of which are enhancements. This issue is about providing a clearer error message without changing the spec.

(In reply to comment #10) > You're talking about two completely separate issues, both of which are > enhancements. This issue is about providing a clearer error message without > changing the spec. Walter has accepted one of the enhancement requests I was talking about (the one about pure functions). I was not aware the second too is an enhancement request, but after re-reading that part of the D specs I see you are right. So technically your error message is correct. But it's hard to keep in memory all the D specs, especially for D newbies. So I suggest a longer and more wordy error message, because this error message is not easy enough to understand. ---------- According to the D specs this code is illegal, is your patch raising an error on it? pure int sqr(int x) { return x * x; } void main() { return sqr(10); } ---------- According to the current D specs this code is correct: int sqr(int x) { return x * x; } void main() { return sqr(10); } But I can't see this as correct, it's bug prone. So bug 3922 was not a dupe, it was an ehancement request! I have to reopen it.

(In reply to comment #10) > (In reply to comment #9) > > > > I don't understand. Elsewhere I have a bug report that asks for a new error > > message if you don't use the result of a pure function. > > But this bug report just asks for a better error message in the situation of > > using "return x;" in a void function. So I expected an error message like: > > > > foo.d(20): Error: a void function can't return 'int'. > > Returning an int expression from a void function is perfectly legal, so long as > the expression has side effects. This is explicitly allowed by the spec. > http://www.digitalmars.com/d/2.0/statement.html#ReturnStatement Yes, but it was agreed on the newsgroup that this is a misfeature. (Andrei was horrified about it). It should be disallowed.

Without this feature, what should happen with lazy void? void lf(lazy void a) { a; } void main() { int i; lf(i++); }

This bug has been fixed with issue 5399. https://github.com/D-Programming-Language/dmd/commit/c942d51c8b1103d5ce4c3dfc03ae77c07c687cd6