Did you test the newer Files.copy? It seems to use a plain char[] forwarding impl too ([1]) vs the sendfile64 of the channel impl ([2]). Also did you try with some Gigs files? IIRC sendfile had a limitation of a few gigas - and we sadly have artifacts of > 4G even if rare. That said if the size is available it can be tested and the optimization used quite easily I guess.
[1] https://github.com/AdoptOpenJDK/openjdk-jdk11/blob/master/src/java.base/unix/native/libnio/fs/UnixCopyFile.c#L53 [2] https://github.com/AdoptOpenJDK/openjdk-jdk11/blob/master/src/java.base/unix/native/libnio/ch/FileChannelImpl.c#L135 Romain Manni-Bucau @rmannibucau <https://twitter.com/rmannibucau> | Blog <https://rmannibucau.metawerx.net/> | Old Blog <http://rmannibucau.wordpress.com> | Github <https://github.com/rmannibucau> | LinkedIn <https://www.linkedin.com/in/rmannibucau> | Book <https://www.packtpub.com/application-development/java-ee-8-high-performance> Le mer. 19 août 2020 à 14:35, Markus KARG <[email protected]> a écrit : > I do not support your judgement. That bug you mention was fixed nine years > ago, so even the most exotic platforms should have adopted the fix > meanwhile. This IS reliable and not more nor less buggy than any other part > of the JRE. What we should do is give intensive tensting of Stefan's PR, > but not overhastily abstain from it due to FUD. As a long time user of NIO > API on Windows and Linux (speaking of years here) I strongly believe that > the performance benefits will definitively outweigh the potential risks. > -Markus > > -----Ursprüngliche Nachricht----- > Von: Jochen Wiedmann [mailto:[email protected]] > Gesendet: Mittwoch, 19. August 2020 13:05 > An: Maven Developers List > Betreff: Re: Why is old IO API used in maven resolver? > > On Wed, Aug 19, 2020 at 11:49 AM Michael Osipov <[email protected]> > wrote: > > > > Am 2020-08-19 um 11:27 schrieb Jochen Wiedmann: > > > Looking at > > > > > > https://bugs.java.com/bugdatabase/view_bug.do?bug_id=6822107 > > > > > > I would very much like us to abstain from every attempt to go into the > > > direction, that Stefan suggests. The consequence could be, that we > > > need to bother with exotic corner cases. Let's stay simple, but > > > reliable. > > > > > > You proposal is not to change the code at all? > > If it's working fine: Yes. > > -- > > Look, that's why there's rules, understand? So that you think before > you break 'em. > > -- (Terry Pratchett, Thief of Time) > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
