I had bookmarked the thread "Management console has stopped working" and would check it every week or two to see what progress was being made. But it seems to have completely disappeared.
What's up?
I made a mistake and deleted that thread after I mistakenly believed I had addressed the problem 100%. In fact, I did fix one crash at startup, but have since seen a separate rare crash that has given me some difficulty debugging. Only rarely do I ever encounter a problem I can't 'just see', and even rarer do I encounter a problem I have difficulty reproducing and debugging. In fact, I can not remember a more frustrating issue. I am sure I will have it resolved soon though, as frankly my entire life is riding on getting it resolved soon. While it doesn't affect everyone, and only affects some people a small portion of the time (99% of cases occurring at startup), I consider it to be a show stopper until fixed.
In the meantime, I certainly do apologize for this annoyance (as no other effects will be seen). I will post updated information here.
Jeremy,
In your post, I count three "mistake*," two "problem," and one each "apologize" and "annoyance."
Don't be too hard on yourself! We all appreciate the fantastic work you do.
LOL, thanks. There is no better destroyer of my own self-confidence and public perception than myself. Still, the fact remains that I must resolve this, no matter what it takes. If I have to, I will rip everything to shreds.
The issue has, and continues to be, that until I get a good debugging session on this crash, or can infer it from trace output, it is very difficult for me to know if it is fixed. There have been several times I thought it was fixed, only to see it on reboot #42 in testing, or just randomly after I'd decided it is surely resolved. That is the problem with not knowing the precise cause, an embarrassing admission indeed. It does not appear to be something excessively simple. In fact, it could ironically be a side-effect of the heavy use of synchronization objects to protect the various data objects referenced by the highly multi-threaded Process Lasso. As with all development, the mother of all evil is complexity ;(
I thought you removed this thread due to OT discussion (began by me, sorry).
Nah, I removed it because I had made myself look like more of an idiot than usual.
I think I have made real progress on this, though am not declaring victory until I complete testing. A new build is definitely coming in the next 18 hours :)
Must be a very rare crash, haven't had one since uh well since the thread. Sometimes the GUI is a bit gooey though. Like it must swim through molasses.
Hmm, I don't know about a gooey interface. It could be the graph - you may want to try hiding it. The graph drawing induces more page faults than any other operation, with the exception of the process icon lookups.
Never have graph active. But since beta 5.0.0.2 it's as responsive as can be. Probably some weirdness on my side. And no problems at 3 minutes delay 8)
I re-enabled some optimizations of the process icon code that I had disabled for diagnostics. This includes asynchronous icon resolution, which should help tremendously.
it did :)
Good, at least something worked ;p.
I have determined the cause of the crash reported by some users in the governor - which I believe is the only one remaining at this time. This was manifested by use of the instance count limits, though was a general problem with the code and some new optimizations.
The initial fix will be in v5.0.0.10 beta, to be released shortly.