Paul Backus
Paul Backus
> * Go ahead and remove `deallocate` and `reallocate` from `GCAllocator` > * Add a `ManagedMemoryAllocator` OSLT with the current unsafe behavior of `GCAllocator` Sure, sounds good. I'll submit a...
> I'm not sure GCAllocator should exist. What is its purpose? Currently? Not much. If this PR goes through? To provide an interface between generic allocator-aware library code (e.g., containers)...
> I'm still trying to figure out what reason one would have to want to do that. Same reasons you'd want to use GC over manual memory management in any...
@atilaneves Ping, anything I can do to help move this and #8555 forward? Would be happy to discuss options at this weekend's online beerconf, if that would help.
@atilaneves Ping.
@RazvanN7 The most granular possible breakdown would be something like: 1. Add `BorrowedRegion`. 2. Refactor `Region` to use `BorrowedRegion` internally. 3. Remove use of `NullAllocator` as a sentinel from `Region`....
@SixthDot Because of the way the calling code uses lookahead, the line in question (error message under `case TOK.auto_:` in `parseFunctionLiteral`) is currently unreachable. I left it in as a...
Auto-tester failure is a segfault in [this static destructor][1] that occurs non-deterministically when running `std.concurrency`'s test suite. Probably a race between the main thread writing to the `__gshared` variable `scheduler`...
@ljmf00 Yes.
Force-pushed to re-run now that the race condition has (hopefully) been fixed.