1 Darwin/MacOSX Support - December 16, 2003
2 =========================================
7 GC_init() MUST be called before calling any other GC functions. This
8 is necessary to properly register segments in dynamic libraries. This
9 call is required even if you code does not use dynamic libraries as the
10 dyld code handles registering all data segments.
12 When your use of the garbage collector is confined to dylibs and you
13 cannot call GC_init() before your libraries' static initializers have
14 run and perhaps called GC_malloc(), create an initialization routine
15 for each library to call GC_init():
18 extern "C" void my_library_init() { GC_init(); }
20 Compile this code into a my_library_init.o, and link it into your
21 dylib. When you link the dylib, pass the -init argument with
22 _my_library_init (e.g. gcc -dynamiclib -o my_library.dylib a.o b.o c.o
23 my_library_init.o -init _my_library_init). This causes
24 my_library_init() to be called before any static initializers, and
25 will initialize the garbage collector properly.
27 Note: It doesn't hurt to call GC_init() more than once, so it's best,
28 if you have an application or set of libraries that all use the
29 garbage collector, to create an initialization routine for each of
30 them that calls GC_init(). Better safe than sorry.
32 The incremental collector is still a bit flaky on darwin. It seems to
33 work reliably with workarounds for a few possible bugs in place however
34 these workaround may not work correctly in all cases. There may also
35 be additional problems that I have not found.
37 Thread-local GC allocation will not work with threads that are not
38 created using the GC-provided override of pthread_create(). Threads
39 created without the GC-provided pthread_create() do not have the
40 necessary data structures in the GC to store this data.
43 Implementation Information
44 ==========================
45 Darwin/MacOSX support is nearly complete. Thread support is reliable on
46 Darwin 6.x (MacOSX 10.2) and there have been reports of success on older
47 Darwin versions (MacOSX 10.1). Shared library support had also been
48 added and the gc can be run from a shared library. There is currently only
49 support for Darwin/PPC although adding x86 support should be trivial.
51 Thread support is implemented in terms of mach thread_suspend and
52 thread_resume calls. These provide a very clean interface to thread
53 suspension. This implementation doesn't rely on pthread_kill so the
54 code works on Darwin < 6.0 (MacOSX 10.1). All the code to stop and
55 start the world is located in darwin_stop_world.c.
57 Since not all uses of the GC enable clients to override pthread_create()
58 before threads have been created, the code for stopping the world has
59 been rewritten to look for threads using Mach kernel calls. Each
60 thread identified in this way is suspended and resumed as above. In
61 addition, since Mach kernel threads do not contain pointers to their
62 stacks, a stack-walking function has been written to find the stack
63 limits. Given an initial stack pointer (for the current thread, a
64 pointer to a stack-allocated local variable will do; for a non-active
65 thread, we grab the value of register 1 (on PowerPC)), it
66 will walk the PPC Mach-O-ABI compliant stack chain until it reaches the
67 top of the stack. This appears to work correctly for GCC-compiled C,
68 C++, Objective-C, and Objective-C++ code, as well as for Java
69 programs that use JNI. If you run code that does not follow the stack
70 layout or stack pointer conventions laid out in the PPC Mach-O ABI,
71 then this will likely crash the garbage collector.
73 The original incremental collector support unfortunatelly no longer works
74 on recent Darwin versions. It also relied on some undocumented kernel
75 structures. Mach, however, does have a very clean interface to exception
76 handing. The current implementation uses Mach's exception handling.
78 Much thanks goes to Andrew Stone, Dietmar Planitzer, Andrew Begel,
79 Jeff Sturm, and Jesse Rosenstock for all their work on the
86 Older Information (Most of this no longer applies to the current code)
87 ======================================================================
89 While the GC should work on MacOS X Server, MacOS X and Darwin, I only tested
91 I've added a PPC assembly version of GC_push_regs(), thus the setjmp() hack is
92 no longer necessary. Incremental collection is supported via mprotect/signal.
93 The current solution isn't really optimal because the signal handler must decode
94 the faulting PPC machine instruction in order to find the correct heap address.
95 Further, it must poke around in the register state which the kernel saved away
96 in some obscure register state structure before it calls the signal handler -
97 needless to say the layout of this structure is no where documented.
98 Threads and dynamic libraries are not yet supported (adding dynamic library
99 support via the low-level dyld API shouldn't be that hard).
101 The original MacOS X port was brought to you by Andrew Stone.
109 Note from Andrew Begel:
111 One more fix to enable gc.a to link successfully into a shared library for
112 MacOS X. You have to add -fno-common to the CFLAGS in the Makefile. MacOSX
113 disallows common symbols in anything that eventually finds its way into a
114 shared library. (I don't completely understand why, but -fno-common seems to
115 work and doesn't mess up the garbage collector's functionality).
119 Jeff Sturm and Jesse Rosenstock provided a patch that adds thread support.
120 GC_MACOSX_THREADS should be defined in the build and in clients. Real
121 dynamic library support is still missing, i.e. dynamic library data segments
122 are still not scanned. Code that stores pointers to the garbage collected
123 heap in statically allocated variables should not reside in a dynamic
124 library. This still doesn't appear to be 100% reliable.
127 Brian Alliet contributed dynamic library support for MacOSX. It could also