stochastic sparc memprotect failure
Greetings! GCL and its reverse dependencies makes use of a garbage
collection algorithm in which pages are marked read-only with
mprotect, segfaults on attempted writes are trapped, pages then
remarked read-write, and finally using the separation of read-only and
writable pages to accelerate collection.
This has worked on sparc for many years, but for the last two or so,
when random failures in the expected behavior are causing memory
corruption. Disabling the algorithm removes all failures. Yet
failures are not fully reproducible, but appear occasionally from run
to run when investigating with gdb.
Camm Maguire email@example.com
"The earth is but one country, and mankind its citizens." -- Baha'u'llah