On 10 December 2011 12:13, Philippe Mouawad <[email protected]> wrote: > I commited another fix by getting child (Downloaders) use their parent . > Tell me if it seems ok for you > > Tests are fine and it looks less impacting for me. > But I agree with best thing would be to make context readonly.
Another approach would be just to copy the required variables to the new thread; in this case I think it's only necessary to copy the source IP - the other settings I think aren't needed for the embedded samples. However, it would be easy to overlook other required variables. > Regards > Philippe > > On Sat, Dec 10, 2011 at 1:10 PM, sebb <[email protected]> wrote: > >> On 10 December 2011 11:55, Philippe Mouawad <[email protected]> >> wrote: >> > I think I understand why there's an issue, by making it >> > InheritableThreadLocal, StandardJMeterEngine root thread will share it >> with >> > children (Thread Group) so big issue. >> >> Good point. >> >> > So it seems the fix will have to be different. >> >> Yes; will need somehow to pre-set the embedded thread context with a >> copy of the original. >> Either clone, or perhaps use a wrapper to provide a read-only copy (to >> save memory and prevent accidents). >> >> > On Sat, Dec 10, 2011 at 12:43 PM, sebb <[email protected]> wrote: >> > >> >> On 10 December 2011 11:16, Philippe Mouawad <[email protected] >> > >> >> wrote: >> >> > My answers below, but there seem to be an issue as tests are failing. >> >> >> >> Yes, strange that adding inheritable has that effect. >> >> >> >> We also need a (batch) unit test to expose the original bug. >> >> >> >> > Regards >> >> > Philippe >> >> > >> >> > On Sat, Dec 10, 2011 at 12:06 PM, sebb <[email protected]> wrote: >> >> > >> >> >> On 10 December 2011 09:25, Philippe Mouawad < >> [email protected] >> >> > >> >> >> wrote: >> >> >> > Hello Sebb, >> >> >> > Do you suggest we fix it That way ? >> >> >> > I Think it should be ok because to make some post processing and >> add >> >> >> > variables to context URL need to be called from réal sampler not >> one >> >> in >> >> >> > parallèl download , am I right in this ? >> >> >> >> >> >> That's what I was thinking too. >> >> >> >> >> >> > Do you see other ways of altering jmetercontext from parallèl >> >> download ? >> >> >> >> >> >> Cookies? Do these still get stored? Cannot remember off-hand. >> >> >> >> >> > We fixed that by cloning CookieManager and merging afterwards. >> >> >> >> Sorry, I meant the optional storing of cookies as variables. >> >> >> >> > >> >> >> >> >> >> > Regards >> >> >> > Philippe >> >> >> > >> >> >> > On Saturday, December 10, 2011, <[email protected]> wrote: >> >> >> >> https://issues.apache.org/bugzilla/show_bug.cgi?id=52310 >> >> >> >> >> >> >> >> --- Comment #3 from Sebb <[email protected]> 2011-12-10 01:00:19 >> UTC >> >> --- >> >> >> >> (In reply to comment #2) >> >> >> >>> The issue is due to the fact the property is cached in >> JMeterContext >> >> >> > which is >> >> >> >>> thread local in Parent thread of Parallel Download threads. >> >> >> >>> Quick fix would be to make ThreadLocal InheritableThreadLocal >> but as >> >> >> >>> JMeterContext is not thread-safe it is not the right fix. >> >> >> >> >> >> >> >> If the context is not mutated, then it should be OK to share. >> >> >> >> >> >> >> >> -- >> >> >> >> Configure bugmail: >> >> >> > https://issues.apache.org/bugzilla/userprefs.cgi?tab=email >> >> >> >> ------- You are receiving this mail because: ------- >> >> >> >> You are the assignee for the bug. >> >> >> >> >> >> >> > >> >> >> > -- >> >> >> > Cordialement. >> >> >> > Philippe Mouawad. >> >> >> >> >> > >> >> > >> >> > >> >> > -- >> >> > Cordialement. >> >> > Philippe Mouawad. >> >> >> > >> > >> > >> > -- >> > Cordialement. >> > Philippe Mouawad. >> > > > > -- > Cordialement. > Philippe Mouawad.
