Re: [hibernate-dev] upgrade to gradle 1.0-m9

I have a different take. If it does not fix anything (and actually causes new problems you report) why make everyone have to upgrade?

On Wed 21 Mar 2012 02:43:27 AM CDT, Strong Liu wrote: > On Mar 21, 2012, at 6:23 AM, Steve Ebersole wrote: > >> GRADLE-2169 (or at least tests hanging on your system) was the main reason to switch to m9. If that's not addressed by m9 then IMO we should not switch. Like I said on IRC all the reports I saw (JIRA and forum) said that that particular thing is still an issue. Curious it only fails on some systems. The next release is supposed to be the first RC, I'd vote we wait for that if possible. One thought I have been having... Is Gradle using its daemon now by default? If so, I wonder if our memory settings, when set environmentally, are not passed along to the daemon. One of the forums posts suggest the underlying cause of these hangs is situations where the "worker process" encounters certain types of errors such as OOM. That would certainly happen trying to run the testsuites. > > no, the daemon is not started by default, it just be considered matured enough. > you still need "--daemon" explicitly to enable it > > GRADLE-2169 maybe not the one i'm getting, at least our tests do not use any security stuff I believe. > > but since my change is ready to push, so, I would suggest you give it a try, if it doesn't break anything (though it doesn't fix my issue either) I would vote we accept it, since with this change, it would make the next upgrade to the CR release easier > >> >> As for name-spacing the manual, the doctype declaration for manual now all use SYSTEM instead of PUBLIC. So even if they cannot be resolved locally, it should never try to resolve them remotely. So not sure why this causes extra time. I would think either it does not load the DTD or it fails. Of course if you already have the work done to convert it to be name-spaced, I'd say push it. > > don't know why, but from my local testing, the dev guide / quickstart docs are built much faster than manual when using dtd > >> >> On Tue 20 Mar 2012 01:11:53 PM CDT, Strong Liu wrote: >>> I have pushed my changes [1] >>> >>> some notes here: >>> >>> * this upgrade does not resolve the GRADLE-2169 issue, my build still hangs (but with "-i" option, it disappears, needs to dig more ) >>> * the gradle-jdocbook change doesn't compatible with previous gradle release (there is a gradle class renamed) >>> * I have to change manual's xml to use namespace instead of the old dtd, jdocbook plugin can't resolve those locally, that's why it takes so long to finish [2] >>> * I haven't created pull request yet, just want to hear feedback first >>> >>> >>> 1. https://github.com/stliu/gradle-jdocbook/tree/upgrade-to-m9 >>> https://github.com/stliu/hibernate-orm/tree/upgrade-to-m9 >>> >>> 2. >>> time gradle :documentation:renderDocBook_manual_en-US_html >>> ----------------------------------------------------------- (mac + m9) >>> Total time: 1 mins 43.315 secs >>> >>> real1m43.546s >>> user1m14.794s >>> sys0m2.919s >>> ----------------------------------------------------------- (fedora + m8a) >>> Total time: 2 mins 33.44 secs >>> >>> real2m33.527s >>> user1m1.912s >>> sys0m1.590s >>> ----------------------------------------------------------- (mac + m8a) >>> Total time: 3 mins 53.628 secs >>> >>> real3m53.798s >>> user1m12.152s >>> sys0m3.037s >>> ------------------------- >>> Best Regards, >>> >>> Strong Liu<stliu at hibernate.org<http://hibernate.org/>> >>> http://about.me/stliu/bio >>> >> >> -- >> steve(a)hibernate.org >> http://hibernate.org > -- steve(a)hibernate.org http://hibernate.org