version 1.1.1.1, 1999/11/27 10:58:32 |
version 1.1.1.2, 2000/12/01 14:48:28 |
Line 37 a thread stack. If you know that you will only be run |
|
Line 37 a thread stack. If you know that you will only be run |
|
or later, it should be possible to fix this by compiling the collector |
or later, it should be possible to fix this by compiling the collector |
with -DSOLARIS23_MPROTECT_BUG_FIXED. |
with -DSOLARIS23_MPROTECT_BUG_FIXED. |
|
|
Jeremy Fitzhardinge points out that there is a problem with the dlopen |
Since 5.0 alpha5, dlopen disables collection temporarily, |
replacement, in that startup code in the library is run while the allocation |
unless USE_PROC_FOR_LIBRARIES is defined. In some unlikely cases, this |
lock is held. This appears to be difficult to fix, since the collector does |
can result in unpleasant heap growth. But it seems better than the |
look at data structures maintained by dlopen, and hence some locking is needed |
race/deadlock issues we had before. |
around the dlopen call. Defining USE_PROC_FOR_LIBRARIES will get address |
|
space layout information from /proc avoiding the dlopen lock. But this has |
|
other disadvanatages, e.g. mmapped files may be scanned. |
|
|
|
If solaris_threads are used on an X86 processor with malloc redirected to |
If solaris_threads are used on an X86 processor with malloc redirected to |
GC_malloc, it is necessary to call GC_thr_init explicitly before forking the |
GC_malloc, it is necessary to call GC_thr_init explicitly before forking the |