• d-bugmail puremagic.com (70/70) Jan 31 2022 https://issues.dlang.org/show_bug.cgi?id=22718

d-bugmail puremagic.com writes:

https://issues.dlang.org/show_bug.cgi?id=22718

          Issue ID: 22718
           Summary: Grammar for case/default statement is
           Product: D
           Version: D2
          Hardware: All
                OS: All
            Status: NEW
          Severity: major
          Priority: P1
         Component: dlang.org
          Assignee: nobody puremagic.com
          Reporter: deadalnix gmail.com

The grammar for case/default statement incorrectly specify DMD's behavior. I do
believe, based on C and C++'s behavior, and D's principle that valid C code
should either compile in D or be an error, but not have different semantic,
that DMD is right, and the grammar is wrong.

See sample code:

import std.stdio;

void foo() {
    writeln(__FUNCTION__);
}

void bar() {
    writeln(__FUNCTION__);
}

void fun(int i) {
    int n = 3;
    switch (i) {
        while (n --> 0)
        case 0:
        case 1:
            foo();
            bar();

        default:
            writeln("default");
    }
}

void main() {
    fun(1);
}


This prints:
example.foo
example.foo
example.foo
example.foo
example.bar
default

As you'd expect using C's grammar. However, current D grammar, as specified
here: https://dlang.org/spec/statement.html#ScopeStatementList tells us that
all statements following the case/default should be parsed as being part of the
case/default statement, unless they are themselves case/default statement. In
the above exemple, this mean that the case 1 ought to be considered to be
outside the loop,as in:

void fun(int i) {
    int n = 3;
    switch (i) {
        while (n --> 0) {
            case 0:
        }
        case 1:
            foo();
            bar();

        default:
            writeln("default");
    }
}

Not that document D's grammar would cause a deluge of bizarre side effects, and
would differs from C. Instead, case/default should be defined in a manner
similar to LabeledStatement. Thankfully, the compiler seems to already do that.

--