Bug 22473 – dmd foreach loops throw exceptions on invalid UTF sequences, use replacementDchar instead

A simple foreach loop: void test(char[] a) { foreach (char c; a) { } } will throw a UtfException if `a` is not a valid UTF string. Instead, it should replace the invalid sequence with replacementDchar. The foreach code is compiled to call druntime/src/rt/aApply/_aApplycd1(), which calls druntime/src/core/internal/utf/decode() which throws the exceptions. replacementDchar is defined in std.utf as `\uFFFD` The reason to effect this change is it is the same problems autodecoding has. It can't be turned off, it throws, and it may allocate with the gc. Oh, and it's slow.

I think the foreach variable type must be dchar. The following program reproduces the issue: void test(char[] a) { foreach (dchar c; a) { } } void main() { char[] a = "\xf0\x90\x28\xbc".dup; test(a); }

I believe making it either wchar or dchar will cause it to try to convert, and throw on failure.

Ali, you're right. I can't believe I made that mistake. I was even thinking about not making that mistake when I wrote the example.

I think this is the same as issue 20134 (except foreach loops instead of Phobos autodecoding, but they should probably behave in the same way).

Actually this is a duplicate of 14519, which describes foreach loops specifically. I'll update the component of that from dmd to druntime (as it is correctly indicated here). *** This issue has been marked as a duplicate of issue 14519 ***