Reply to thread

Message

<blockquote data-quote="xppx" data-source="post: 1086065" data-attributes="member: 55884">Worked again late today...Did test the 1.6.0-pre build. See result.Just a thought - most probably a stupid question - but how do you manage synchronization in MP? As developer I would look for methods that are running in parallel and do not respect an order when they actually should: the cleanup might take too long (which might cover another issue to be resolved) and the initiation of the next video thread perhaps doesn't wait until the cleanup is finished... (if both methods were synchronized against the default lock manager, they would be slower but never run at the same time and thus avoid a racing condition) - typical side effect is that errors that might appear or not depending on the hardware (number of CPU's, speed) and the system (number of available threads) and the software (each compile might change the behavior a bit). And we see the error appear on the 3rd video for 1.6 in debug, 4th video for 1.7 and 1.8pre in debug, and 2nd video in normal mode...What do you think?Result: still error[code]MP 1.6 pre - Commit: SHA-1: 5f3b5d1c7e8f1c5bc9d453043270b726d19059ecRefresh rate changer disabledResult: Error after start of 3rd video (in 1.7 and 1.8 this happens on the 4th in debug mode and on the 2nd in normal mode)[/code]</blockquote>

[QUOTE="xppx, post: 1086065, member: 55884"] Worked again late today... Did test the 1.6.0-pre build. See result. Just a thought - most probably a stupid question - but how do you manage synchronization in MP? As developer I would look for methods that are running in parallel and do not respect an order when they actually should: the cleanup might take too long (which might cover another issue to be resolved) and the initiation of the next video thread perhaps doesn't wait until the cleanup is finished... (if both methods were synchronized against the default lock manager, they would be slower but never run at the same time and thus avoid a racing condition) - typical side effect is that errors that might appear or not depending on the hardware (number of CPU's, speed) and the system (number of available threads) and the software (each compile might change the behavior a bit). And we see the error appear on the 3rd video for 1.6 in debug, 4th video for 1.7 and 1.8pre in debug, and 2nd video in normal mode... What do you think? [B]Result: still error[/B] [code]MP 1.6 pre - Commit: SHA-1: 5f3b5d1c7e8f1c5bc9d453043270b726d19059ec Refresh rate changer disabled Result: Error after start of 3rd video (in 1.7 and 1.8 this happens on the 4th in debug mode and on the 2nd in normal mode)[/code] [/QUOTE]

Verification