Bug 8507 – (Regression 2.060) rdmd hangs and mutex failures

(There's no rdmd component, so...) I don't know if this bug exists on other platforms, but on Linux (x86_64) the rdmd in 2.060 eventually hangs (and in some cases even crashes due to mutex failures). Given a test.d like this: import std.stdio; void main() { writeln("Hello World"); } Try running: for x in {0..9999}; do echo iteration $x; rdmd test.d; done This should yield several mutex errors, and eventually make one of the rdmd invocations hang. This is a bit of a pain since my program's test suite relies on rdmd to run the suite.

I was also experiencing similar issues on my previous system (running Windows Vista x64). The problem went away after I changed the line "if (false)" to "if (true)", thus disabling the multi-threaded stat-ing. However, I can't reproduce the problem on my current system (Windows 7 x64). I think we should just remove the multithreaded stat until the underlying bug can be fixed. AFAIK there is no recorded measurement that it provides an actual performance benefit.

Works fine on my Linux x86-64 system.

I've measured MT stat a while ago (i.e. when I introduced it) and at least on Linux it helps. Can this bug be reproduced reliably on Windows?

Works for me on Windows (dmd 2.070.1)

Thanks. I'll close this, feel free to reopen with new evidence.