* Added Microsoft windbg.exe Win32 debugger. * Removed std.recls, instead use http://synesis.com.au/software/recls/. * Incorporated Dave Fladebo's changes: o std.regexp speedups. o Fixed a bug in pauseHandler whereby the stackTop was being overwritten by subsequently resumed threads causing a crash in gcx.mark(). (The fix was to just move sem_post(&flagSuspend); to right before the sigsuspend loop). o Moved linux semaphore declarations to std/c/linux/linux.d o Changed void wait(uint milliseconds){} to be a timed wait function instead of just a stub for void wait() {} on linux (it is comparable to the Win32 Thread method now). This is implemented with pthread mutexes and conditions and the extra overhead to initialize phobos Threads does not result in slower executables (Threads, GC and startup time are not effected). o Changed the allThreads static array size to the same as Windows. This in conjunction w/ the stack size change will allow the creation of more than 100 threads on linux for better compatibility w/ the Win32 version. o Added optional stack size parameter to constructors. Since the stackBottom is static when the threads are created and the stackTop is dynamic based on the stack pointer (or context for Win32), allowing the threads to be created with a non-default stack size shouldn't interfere with the GC. This adds little overhead to either the Win32 or Linux Thread implementations.
The 'gdc-Ego' branch of GDC is a highly experimental branch of GDC with patches by Gregor Richards, using the the DMD 0.156 frontend. Furthermore, it has fixes to several bugs that were in GDC 0.17 which caused inconsistencies between DMD and GDC. It's experimental and has known bugs, but should allow you to compile D software that is only compatible with the latest DMD.
The gdc-Ego branch of GDC is a very experimental branch of GDC that puts the patch of Gregor Richards. The front end of DMD 0.156 is used. In addition, the bug of shoes that want to cause the contradiction between DMD and GDC that exists in GDC0.17 is corrected. The software written in only latest DMD and an incompatible D language should be able to be compiled though the bug of knowing exists experimentally, too.
/usr/include/mach/ppc/vm_types.h:86: error: conflicting types for `typedef int integer_t' ../../gcc/d/dmd/mars.h:146: error: previous declaration as `typedef long long unsigned int integer_t'