Psst, check his sig.

Very true. Some have lotteries and whatnot occasionally but not regularly.

Well, not just p2pool but any non-pps pool will be making you more money after you've found the block. So proportional, DGM, slush scoring, smpps etc etc all will increase their returns to you as a result of you finding it.

Right, you shouldn't need dummy plugs unless you have a driver before about 11.6 or something. I still suspect your xorg config. is wrong, or you're not really exporting the display variable. Also it's worth noting people have had looots of trouble with recent drivers and last good was something like 12.8. Can you show us the actual output from cgminer -n ?

With stratum you already roll nonce2 which has a variable number of bytes... currently it has 4 bytes which gives you a heck of lot more range than just 7200 rolls... There is no need to roll the time on top of that, and in fact the stratum spec says you can only modify the time to be consistent with the real time (i.e. increment by one every second). That still gives you 2^32 rolls per second with stratum. So 2^32 unique work units per second, each with 2^32 nonce like previously, and there is scope to actually increase the number of bits of nonce2 as well. It will be nice if the block times are more accurate as we move away from rolling the time as a hack to make more work, and stratum makes it possible.

Actually sounds like it's still only mining on the one  GPU, just splitting it into two. What does 'cgminer -n' say?

By the way I spotted that it is still possible for a few untracked shares to rarely show up, but for the most part it is only a cosmetic issue and is corrected in the current git tree.

Yes, being a local proxy it responds quickly which is why you can reproduce it. Try 2.8.7 please.

Hold on while I respin the 2.8.6 release as 2.8.7 ... it hasn't been out that long  

Pretend you didn't see anything...

Actually come to think of it, perhaps the results are coming back before the database entry is even being created. Hmm

That means the stratum pool you are mining with is returning "accepted" messages without the same ID that cgminer sent the share with, so cgminer doesn't know which shares the pool is accepting. The communication is all done asynchronously and cgminer keeps a database of sent shares so that when it gets a response from the pool it can tell you which share it was.

Not quite... it actually accepts diff 1 shares and tests them all against its internal variable difficulty, and still operates off the getwork protocol. In its current form I'm pretty sure it would consume a lot of CPU trying to keep ASICs busy since it already consumes quite a lot of CPU without this added load.

New versions -> 2.8.7, 29th October 2012

Mainly stratum fixes for windows plus minor tweaks and cosmetic changes.


Human readable changelog

After much wrestling I've discovered that windows happily sends stuff into sockets for a long time after they're dead and finally found a workaround that would realise it couldn't send them. I'm assuming the crashes people were seeing on windows are related to overflows with this phenomenon and hopefully this will improve stratum reliability further on windows.
Pool timeouts on unresponsive pools with stratum should be picked up sooner.
Failure to submit shares with stratum now also registers as an RF remote failure.
Kano tracked down a longstanding memory leak which was small but did add up over time when GPU mining.
There is a new "compact" display mode you can specify with --compact or enable it in the menu during runtime which only shows the summary stats and not a line for each device, in case you have heaps of devices and they don't fit on screen.
The best share difficulty is now displayed in the status window at the top, and in the summary on exiting.
Some details have been trimmed from the pool listing in the status window in the interest of screen real estate efficiency.


Full changelog

- Fail on select() failing in stratum thread without needing to attempt
recv_line.
- Add share to stratum database before sending it again in case we get a
response from the pool before it's added.
- Shorten the initiate stratum connect timeout to 30 seconds.
- Shorten the stratum timeout on read to 90 seconds to detect unresponsive pool.
- Display best share difficulty on exit.
- Make stratum socket fail more robust on windows by disabling the send buffer.
- Reuse the same curl handle forcing a new connection instead of risking
derefencing.
- Add information about submission failure to stratum send.
- Only add stratum share to database if we succeeded in submitting it, with a
debug output saying it succeeded.
- Use keepalive with stratum sockets to improve its ability to detect broken
connections.
- Show only the URL in the status bar to avoid long prefixes making for extra
long lines.
- Display compact status in menu and update README to reflect current menu
entries.
- Add a compact display mode that does not list per device statistics in the
status window.
- Add blank spaces after best share displayed.
- Round a few static string arrays up to 4 byte boundaries for ARM.
- Display best share diff for scrypt as well.
- Show the best diff share as "best share" and add info to the README.
- Display the best diff share submitted so far.
- Redundant check.
- The work struct pointer in struct pc_data in findnonce is never freed yet
there is no need to allocate it separately so make struct work a static part of
the struct pc_data.

With stratum mining that's whatever return code the server has given you for the rejection. That one usually means it has asked you to start on new work and that share is from the old work. i.e. stale.

It does, from the work.

Already underway

This is new in the display of the next version:

I got a share that registered a diff of 7.58M 

That was a block solve for mtred since current difficulty is just over 3 million.

I wasn't saying that was necessarily the cause of the crash, but it's worth noting that you had an overheat anyway.

REST is usually due to an overheat event, unlike SICK and DEAD.

Right.