Egor Pasko wrote:
On the 0x39D day of Apache Harmony Yang Paulex wrote:
3.The most important thing is we can't debug our program, which is a
magnificent and complex program, although you can debug a small
smoothly.
Not sure what VM you were using, but I believe there are many guys here
happy to help fixing the problems :).
probably, Junguo meant that debugging is very slow with DRLVM?
In fact, DRLVM should be slower when debugging because many tricky
JIT optimizations are turned off in this mode (for stability reasons).
There were two (two on Linux, one on all platforms) regressions in
debugging code that could easily deadlock debugged process. These are
HARMONY-1575 and HARMONY-5213. The 2nd one I fixed just an hour ago.
These two bugs happened after introducing additional synchronization
(HARMONY-4862, revision 594177) to protect necessary data in
multithreaded programs and generally improve debugging stability. But
new synchronization code happened to have bugs and therefore caused
deadlocks.
Now that known regressions are fixed I hope debugging quality is better
and I would really like if Junguo tries it again. The revision with both
bugs fixed is 599381.
--
Gregory
