LSB conf call notes for 2008-10-15

Printer-friendly versionAlexey, Vladimir LSB 4.0. Jeff: builders were the biggest problem. Mats: ia32 issue that was on IRC; some of the builds for that architecture weren't built off the tags, and are too old. Jeff: any other archs? Mats: ok on x86-64; don't have results on anything else. Darren: what's the problem? Mats: dtk-manager doesn't pick up on the cairo test journal. Darren: Jiri is running tests with the beta now; should be able to report if it's a problem. Jeff: can those results be sent to us? Darren: not a problem, except workload. Jeff: can do automatically. Jeff: new autotest results page. Would like to see people contribute results, request new reports. Jeff: any worrisome results? Jeff: back to builders. Clearly, it's too fragile. What to do? Mats: question is too complicated; it doesn't work, but what doesn't work is an issue. Jeff: start from scratch? Russ: could also refactor. Mats: we have a sense of where the problems are. Uploads fail much of the time. Another weakness: first phase is to build the development environment, and have it replace itself. The sequence is tricky, and can fail. Russ: tinderbox-type environment? Cascading change issue; might be able to catch these problems earlier. Jeff: also, keeping chroots around is fragile; keeping pristine tarballs for builds is better. Mats: should kick off a project to improve this. Jeff: builder doesn't probably take advantage of new SDK capabilities. Stew: that part doesn't fail as much. Mats: would be cool to use virtual machines as builders; could do copy-on-write off a pristine builder and throw changes away. Jeff: how do we kick off the project? Use an architecture as a guinea pig? Mats: if we can build without pushing, then we can test without forcing changes into production. Russ: conflating chroots with build instance; using copy-on-write. Where is the proposal? Jeff: where is your stuff? Russ: old code is at ftp://ftp.owlriver.com/pub/mirror/ORC/ORCrebuild/ Personally uses mezzanine, centos uses mach. Always build packages, which may be a discipline change. Jeff: also, don't have a good virtual machine situation for some archs. Russ: IBM demonstrated something; ia64 is almost there, and s390 is there. Kay: have a program for creating VMs for ChipHopper people. It's a bit complicated. Russ: can use snapshots of VMs for creating new ones, instead of "stock" images or new installs. Jeff: schedule. Could gain time. Who? Michael and I? Mats: would like to volunteer, but lots on plate; in particular, the 3.2 refresh. Would like to be involved. Jeff: could make our main job, but pull Mats in for discrete tasks. Kay: update on Java stuff? Jeff: no progress in a month. Kay: even if someone packages Java with their app, the fact that there's no LSB-certified JDK is a problem. Russ: problem with testing. End-of-lifing "100% Pure Java" is very troubling. No good test situation. Robert: should push on both ends: Java in LSB, and LSB-certified Java. Kay: not given up on certified Java side. Jeff: need to do the vote yet on trial-use, but we're pretty sure Java isn't ready. Robert: maybe after the 4.0 hump. Russ: and making sure we have application testing is important. Russ: concerned that the vote on trial-use status for Java will not happen the way we expect. Jeff: have been non-committal, but most of the people involved have been agreeing that Java should be trial use. Russ: yes, but what about ballot surprises? Jeff: will probably put out a position paper on the issues to vote on, and will include the technical recommendation of the LSB workgroup and Sun.