Peter, to me it looks like the issue is _not_ in resolver, it looks more like it is in Maven itself. Could you try out the latest master build of Maven? I can build it for you and share it with you, if you are ok with that. Or, just check out https://github.com/apache/maven and build it?
Thanks T On Sat, Dec 21, 2024 at 2:37 PM Peter Teixeira <pteixe...@hubspot.com.invalid> wrote: > > Hello! Thanks for taking the time to look into this. I'd definitely be > willing to believe that this > is strange going on with these builds, but I'll keep working on getting an > actual reproducer set up. > > I didn't have any luck with setting the http protocol back to 2. I *did* > wind up running Maven with > JFR (via `MAVEN_OPTS="$MAVEN_OPTS > -XX:StartFlightRecording=filename=recording.jfr" mvn dependency:tree`) > to see if I could collect some timing data, in case you think that'd be > worth looking into. > > recording-rc1.jfr > <https://drive.google.com/open?id=1fyzhR5Avgrg15sFpv5Pt3z5fiuubTLpx> > > recording-beta5.jfr > <https://drive.google.com/open?id=1VtoKbI8WNTC_JDZgssG-AGAv1jUeKQg4> > > > Thanks for your help! > Peter Teixeira > > On Sat, Dec 21, 2024 at 7:42 AM Tamás Cservenák <ta...@cservenak.net> wrote: > > > Found a slightly bigger tree: > > https://gist.github.com/cstamas/c255de93292ecbf1f6f0597ac58624a8 > > <https://gist.github.com/cstamas/c255de93292ecbf1f6f0597ac58624a8> > > > > Still no noticeable diff, especially not in order of magnitude like > > seconds vs minutes... > > > > Something else is going on... > > > > Thanks > > T > > > > On Sat, Dec 21, 2024 at 1:34 PM Tamás Cservenák <ta...@cservenak.net> > > wrote: > > > > > > Still shooting in the dark... > > > but just to rule out any mischief in any plugin, > > > I used my own code, and ran it with 3.9.9 and 4.0.0-rc-1 on a project > > > I could find, that has fairly big(ger) tree: > > > > > > https://gist.github.com/cstamas/c2751bd8c18eb553f3c38d839c60fc67 > > <https://gist.github.com/cstamas/c2751bd8c18eb553f3c38d839c60fc67> > > > > > > As you see, both finished in almost exactly the same duration... > > > So this is not a resolver collection for sure. > > > > > > Thanks > > > T > > > > > > On Sat, Dec 21, 2024 at 1:24 PM Tamás Cservenák <ta...@cservenak.net> > > wrote: > > > > > > > > Hi Peter, > > > > > > > > one obvious thing comes to my mind: > > > > Maven 4.0.0-rc-1 uses Resolver 2.0.4, not 2.0.3. > > > > > > > > And one big change in Resolver 2.0.4 was backing from HTTP/2 protocol > > > > to HTTP/1.1 transport for default JDK > > > > (that, in the scenario of getting a lot of small data like Maven > > > > metadata are, does matter). > > > > > > > > Can you rerun your commands with added following parameter: > > > > -Daether.transport.jdk.httpVersion=HTTP_2 > > > > > > > > This will re-enable HTTP/2 for JDK transport. > > > > > > > > Thanks > > > > T > > > > > > > > On Sat, Dec 21, 2024 at 3:06 AM Peter Teixeira > > > > <pteixe...@hubspot.com.invalid> wrote: > > > > > > > > > > Hello, > > > > > > > > > > First, I sincerely apologize that I don't have a reproducer for > > this, but I > > > > > did want to say something before the full release of Maven 4. > > > > > > > > > > It seems like for certain dependency tree setups, dependency > > resolution got > > > > > much much slower between Maven 4-beta-5 and Maven 4-rc-1. I > > specifically > > > > > think this was introduced in maven-resolver 2.0.3, and building Maven > > > > > 4-rc-1 patching back to maven-resolver 2.0.2 restores the behavior I > > was > > > > > seeing before. > > > > > > > > > > I have a gist > > > > > < > > https://gist.github.com/PtrTeixeira/b7ad68d2e58c6a2277206b2b6328dc16 > > <https://gist.github.com/PtrTeixeira/b7ad68d2e58c6a2277206b2b6328dc16>> > > of > > > > > the behavior that I'm seeing; for Maven 3 & Maven 4-beta-5, running > > `mvn > > > > > dependency:tree` on this module takes seconds. For Maven 4-rc-1 and > > Maven > > > > > 4-rc-2, it takes over a minute. > > > > > > > > > > For some background context, at work almost all libraries are used at > > > > > `1.0-SNAPSHOT` versions and are managed in a single parent-pom. I'm > > seeing > > > > > this behavior across many repositories at work, so it's not something > > > > > specific to this module. While trying to reproduce this, I did > > establish > > > > > that it doesn't seem to be specifically a result of the depth of the > > > > > dependency tree or the total number of dependencies in the > > dependency tree. > > > > > > > > > > Please let me know if there's any other information that I can give > > you to > > > > > help identify this or track this down; I know this isn't a lot of > > > > > information to go on, but this is enough of a performance change > > that I > > > > > wanted to say something before Maven 4 gets fully released. > > > > > > > > > > Thanks, > > > > > Peter Teixeira > > > > > > > > > > -- > > > > > We're committed to your privacy. HubSpot utilizes both public > > information > > > > > and/or the information you provide to us to contact you about our > > relevant > > > > > content, products, and services. You may unsubscribe > > > > > <mailto:opt-...@hubspot.com?subject=unsubscribe&body=unsubscribe> > > from > > > > > these communications at any time. For more information, check out our > > > > > Privacy Policy <https://legal.hubspot.com/privacy-policy>. > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: users-unsubscr...@maven.apache.org > > For additional commands, e-mail: users-h...@maven.apache.org > > > > -- > We're committed to your privacy. HubSpot utilizes both public information > and/or the information you provide to us to contact you about our relevant > content, products, and services. You may unsubscribe > <mailto:opt-...@hubspot.com?subject=unsubscribe&body=unsubscribe> from > these communications at any time. For more information, check out our > Privacy Policy <https://legal.hubspot.com/privacy-policy>. --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@maven.apache.org For additional commands, e-mail: users-h...@maven.apache.org
