H. S. Teoh
H. S. Teoh
I've managed to minimize the code: http://d.puremagic.com/issues/show_bug.cgi?id=9352
Actually, I wonder if the delegate problem can be worked around by declaring .old as a _local variable_ in the ctor (to cause the compiler to allocate it on the...
Whoa. This is so many levels of wrong... but fortunately, it appears that this has been fixed in the latest git HEAD. I tested your code in my environment, and...
Honestly I don't see what this would actually accomplish in practice. What would this achieve that you can't already achieve with launching `dmd` via `std.process` to create a shared library,...
I highly recommend Adam's JNI library. I'd advocate for it to be officially adopted, if at all possible. If not, I'd highly recommend borrowing from his design. He has made...
This is an abuse of `@trusted`. Since `getenv` and `putenv` are not thread-safe, `@trusted` is unwarranted here unless we implement locking.
+1, kill it with fire. It's not only useless, but harmful to code that relies on well-defined behaviour. Let dmd continue with whatever mistakes it wants, we can do better...
Problem is caused by sformat not interpreting surrogate sequences, so it feed it to the encoding function one code unit at a time, but the latter enforces no unpaired surrogates.
What about making expandTilde a template function, so that the linking to NSS, etc., only happens if the user actually calls the function?
Reduced code: ``` import std.array : appender; struct T { string a; alias a this; } void main() { auto t = [ T("Ж") ]; // any non-ASCII char will...