On 15 May 2012 19:50, Milamber <[email protected]> wrote: > > > Le 15/05/2012 16:07, sebb a ecrit : >> On 14 May 2012 21:33, Milamber <[email protected]> wrote: >> >>> >>> Le 13/05/2012 23:10, sebb a ecrit : >>> >>>> On 13 May 2012 21:25, Milamber <[email protected]> wrote: >>>> >>>> >>>>> Hello, >>>>> >>>>> I thinks that is better to wait the release after JMeter 2.7 to upgrade >>>>> HttpComponents HttpCore to 4.2 (currently 4.1.4). >>>>> http://mail-archives.apache.org/mod_mbox/www-announce/201205.mbox/%3C1336239139.2137.34.camel%40ubuntu%3E >>>>> >>>>> >>>> Not sure why you suggest waiting. >>>> Though of course we need to check that it is suitable for use with the >>>> current HttpClient release. >>>> >>>> >>> Because, perhaps, a httpcore 4.2 update may introduce some hidden bug >>> just before the release. But you are right, the update can be made >>> before release with some checks. I can do this if your are ok. >>> >>> >>>> >>>>> On http://www.apache.org/dist/httpcomponents/httpcore/RELEASE_NOTES.txt >>>>> "new asynchronous client and server side protocol handlers", seems >>>>> important for the Parallel Controller discussion. >>>>> >>>>> >>>> I doubt we need to move to aynch. IO for that. >>>> After all, we already support parallel requests in different threads. >>>> >>>> >>> >>> Basic async request is used in HTTPSamplerBase for parallel download of >>> embedded resources. >>> >> Huh? >> >> AFAICT, we use the standard HTTPSampler sample() method in a separate thread. >> >> We don't use Asynch I/O. >> > > Perhaps, I'm confused in my mind. > When I've developed the parallel download of embedded resources, I have > used java.util.concurrent.Future object, and the javadoc says "A Future > represents the result of an asynchronous computation". > I believes that the result it's a Async HTTPSampler in HTTPSamplerBase, > and a I/O (input/ouput) in my mind may be a network I/O.
Yes, the I/O can be network I/O, but it is still synchronous; it uses just the same code as in the main HTTP samplers. The Asynch I/O provided by HttpCore is completely different; it uses NIO. > Milamber > >> >>> Perhaps that is a opportunity to study the async >>> httpclient api for the parallel controller discussion (or embedded >>> resources) >>> >> Perhaps, but from what I have read on the HC mailing list it's only >> really necessary for very high numbers of parallel requests. >> >> >>> Milamber >>> >>> >>>> >>>>> With Commons IO 2.3, Java 1.6 is needed. >>>>> http://mail-archives.apache.org/mod_mbox/www-announce/201204.mbox/%3CCACZkXPznzeiMdT%3DU%3DcLR16G2CKJJUiGzrt501rBREk3aZYxq6A%40mail.gmail.com%3E >>>>> >>>>> >>>> We should only upgrade to IO 2.3 once we start needing 1.6 elsewhere, >>>> rather than being forced to use 1.6 because of IO 2.3. >>>> I don't think we make extensive use of Commons IO, so there is no >>>> particular need to update it. >>>> >>>> >>>> >>>>> After JMeter 2.7 release, I will start the Java 1.6 conversion >>>>> (@override tags, System Tray Icon, etc.) >>>>> >>>>> >>>> Yes. >>>> >>>> >>>> >>>>> Any comments are welcome. >>>>> >>>>> Milamber >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>> >>> >> >
