Le 27/03/2018 à 08:07, Stefan Seelmann a écrit : > On 03/27/2018 12:45 AM, Emmanuel Lecharny wrote: >> >> Yeah, I really think it's the problem. >> >> I have committed a fix (using 30000L instead of 30L) because, anyway, it >> does not make any sense to use such a short timeout. > > Yes, that was clearly the cause, I don't know why that was 30ms, maybe > it was seconds before?
Most certainly. I would assume that it was a mistake in the API design - using seconds instead of milliseconds - that got changed to stick with JNDI timeout that uses milliseconds. But anyway, it's not really problematic, and it went unoticed until I ran the tests on my overloaded computer (I'm pretty sure I never had the problem before, but I wasn't running a VM using 4GB of mem, beside *many* other things, like 2 eclipse instances, Studio running aside, FireFox with 200+ tabs opened, ThunderBird, LibreOffice... This is what you get when you have 16Gb of available RAM and feel comfy never stopping any app ;-) That being said, what took me long was to set up the test env, and like I said, I had to doom my workspace almost completely and reload the API, ApacheDS and Studio - with the subsequent rebuilds - to get the test working :-) Once done, it took me like 15 mins to spot the issue. Thanks Stefan ! -- Emmanuel Lecharny Symas.com directory.apache.org
