On 6/25/07, Daniel Frampton <email@example.com> wrote:
Releasing more often is a better idea so +1 to that.
I also suggest that this release conclude the formal stability drive,
and we begin working on both performance and key features we want for
the next release. Any thoughts?
As for the rest I would really like to see two issues addressed but I am not going to have any time to look into them so feel free to ignore.
* Invalid CMID when walking stack (i.e.
I suspect the size for a frame we are walking is off by a word or two. Thus we occasionally get RVM-87 but I suspect it is also the cause of some of the remaining gc map bugs (if you look at faults the tib address often have values that could be CMIDs). So in theory we should be able to add some debugging into thread scanning and if it gets invalid CMID then search a little up and a little down until a possible CMID is found and print out name of corresponding methods before dumping. Of course this is all just me guess withut having any chance to investigate.
I would like to see the described temporary fix implemented just so we are not silently ignoring corruptions. (Although a full fix would be grand).
Other than that I think the system has improved dramatically and should be able to keep moving with new features !