RE: Re: Fast/Slow polling
| Newsgroups | gmane.comp.emulators.hercules390.general |
|---|---|
| Message-ID | <[email protected]> |
Jürgen Winkelmann wrote: [...] > and my second suspicion is that you are running on a platform > that behaves timing wise differently than the classics (Windows, > Intel Linux), like Mac OS X. I believe the periodic "false" slow->fast->slow transitions that occurs during normal operation (each of which triggers a pair of HHC02917 messages) is a natural side effect of "Timer coalescing" logic in the Windows kernel wherein timer related wakeups (which I'm guessing also includes select() timeouts) don't occur exactly on time, thereby misleading Hercules into erroneously believing a IND$FILE transfer is starting, causing it to switch from slow to fast polling, only to switch back to slow polling again a very short time later, thinking the IND$FILE transfer has ended. Since there's nothing we can do to control the timing (delay) of the when we're woken up by the kernel, there's currently nothing we can do (with the current design) to prevent such "false" messages from periodically occurring during normal operation (other than to suppress them altogether as I've done in the next release). -- "Fish" (David B. Trout) Software Development Laboratories http://www.softdevlabs.com mail: [email protected]