1) We need to know more.

2) This version has improved scheduling, and new restrictions. Suspending work units is perfectly normal, I would have to see specific instances to determine if there is a real problem.

3) Again, the precise error is needed.

What do you mean, "explorer is shut down"?

Please will you help improve BOINC by posting the precise errors, or getting the affected member(s) to do so? The issues you raise don't match any I'm aware of that were seen during beta testing.

Thank you.

1. I've encountered once on a machine which is still on 5.10.5, while it was just with NASA TV in WMP 10 on screen, WXP-MM up to snuff (C2D). The agent and explorer were unloaded, but could recover from it by the Taskmanager fire up (ctrl-alt-del) and starting explorer. This has been reported a few times on BOINC DEV forum also, but only for 5.10.7.

The daemon and sciences had been unloaded as well, but recovered at checkpoints.

2. Had that too and reported on BOINC dev forum with screenshots. Had a batch of 8 FAAH which were tried 1 by 1 and after the last continued on the 5th in the queue. With LIM active had 8x96mb sitting in Ram, then 7, then 6 until it finished the lot. It looked like it was seeking for the shortest job. Seen this a number of times now explicitly, where the GC jobs (short) get priority over FAAH (medium long) and finally HPF2 (by far the longest in this cycle). (P4HT+C2D). No response on BOINC dev forum so far.

3. Not seen at all and been testing since the first 5.10.x release, this side, and La Bella Paese has been under a persistent internet attack for a number of days now. It was today reported on the BOINC dev Forum, with the seeming workaround being to configure a standard connection. http://boinc.berkeley.edu/dev/forum_thread.php?id=1914

The benchmark is rock solid and only uses for consistency Core 0 (and duration cut in half ;O).

Added: @Movieman, WCG had no powers over what's on the general downloads page over at Berkeley.... they can suggest! I like the panic inducing title 'Lots of problems'.... maybe Lot's of people having same 3 problems?

and why WCG is not recommending to use 5.10.7 yet, is because already there is

5.10.8: Development version (Win98+, Win2000+) (MAY BE UNSTABLE - USE ONLY FOR TESTING)

and i happened to gloss over the linux version and it still(?) says:

Linux/x86 details
5.10.7: Development version (standard GUI) (MAY BE UNSTABLE - USE ONLY FOR TESTING)

Related to this: be aware WCG has a minor bug in their device profile page. Allowed memory may be shown as 100% when in fact it isn't (the agent default is in effect). Saving your profile saves the value properly, and updating your agent will force the client to use the correct value.

This bug has been reported, and will hopefully be fixed soon.

There is a problem with the BOINC 5.10.x builds relating to proxies: it doesn't work at all with SOCKS proxies. I don't know about HTTP personally because, since it can't do NTLM, it won't work with our ISA Server Firewall anyway (hence the need for the SOCKS proxy), but there was another report of HTTP and SOCKS failure over at the BOINC forums. I have duplicated the problem on Vista, Vista x64, and XP Professional SP2. I would consider that a fairly major bug: essentially anyone with networks that require proxy use to communicate with the outside world are now hosed if they try 5.10.x. I have opened a ticket in the Trac system and posted to the BOINC forums but so far it seems to be receiving little attention. I'm sure they have bigger fish to fry at the moment though.