[lsb-discuss] LSB conf call notes for 2008-03-26
jeff at licquia.org
Wed Mar 26 09:29:56 PDT 2008
Attendees: Jeff, Robert, Sam, Jesper, Carlos, Mats, Stew, Kay Tate,
Darren, Marvin, Russ Herrold, Alan Clark, Ted, Alexey, George Kraft.
Project planning. Jeff: plan is that LSB 4.0 tasks will either be on
the LSB Project Management list, or a bug in Bugzilla with the proper
target version set. Ted: late on lists of interfaces and libraries that
would enable large numbers of ISV applications. Did we ever get a
mapping on the ChipHopper apps? Kay: Alexey was going to change the
names of the apps in the Navigator based on info from CH. Ted: wanted
to make sure we could prioritize the libs used by the CH apps. Kay:
will bring a presentation to the F2F, getting some additional info from
some of their ISVs.
Ted: lots of wishlist items in ProjectPlan40; which have been turned
into project pages/bugs? Jeff: added multi-version tools, ongoing
database reimports. Mats: multi-version would be a very good thing for
reducing maintenance burden. Ted: has ISPRAS looked at build tools?
Mats: had reported they wanted to do it. Ted: bug on build tools?
Jeff: 1965. Mats: the checkers are mostly multi-version, all except for
JVM. Ted: Sun guy? Jeff: had an internal strategy meeting last week;
hadn't heard how that went. May be on the call next week. Ted: reach
out to others, possibly Ian.
Ted: Use of syscall is a bit problematic; maybe can add syscall to help
Solaris compatibility in return for help on the Java side. Problems:
what to do to make a LSB-compliant JRE, and what do we do to make it
easy for non-Linux systems? (syscall, /proc, /sys) Jeff: could be a
backward-compatibility thing; use new kernel features conditionally via
syscall instead of linking the libc symbols. Ted: appchk could be a
problem. Jeff: could use same approach as dlopen: warn, app vendors are
responsible for fixing apps that use it wrongly. Ted: also, syscall
interface differs between archs, which makes portability between archs
very difficult. Jeff: would be a show-stopper in his book.
George: could syscall info be put in the Navigator? Mats: yes, should
be done; commenting in progress. Ted: do we have a process to save
things like comments? Mats: not a formal process; the database reload
process includes workarounds to ensure they aren't dropped. Comments go
into a separate table which we should be backing up. Mats: Novell apps
among those using syscall(). Darren: will figure out why. Kay will do
the same for IBM/CH apps. Jeff: bug/project? Mats: not at present;
should probably set up a project. Ted: should set this up as a project.
Ted: other stuff on ProjectPlan40? Cairo is starting to be used. Mats:
GTK+ 2.8 and greater require Cairo. Ted: is Cairo ready--docs, etc.?
Mats: GTK+ family is pretty well documented; Cairo may be a little less
so. Jeff: had talked to Behdad. Ted: should document on the wiki page,
thought someone else might be important to pull in. Mats: old desktop
project did have a page; should look over to see if there's any
interesting info to report there.
LSB RPM uplift. Ted: important? Mats: yes. Kay: yes, our spec is very
Mats: Cairo is well-documented, including deltas. George: test suite?
Mats doesn't know.
Jeff: we should definitely uplift RPM, or consider dropping, and
dropping would be very unpopular with a lot of folks. Ted: can we build
LSB RPMs? Mats: with the caveat that many tags are just ignored, yes.
Alexey: appchk for POSIX shell scripts. Darren: should also consider
adding bash directly. Alexey: agree. Ted: using #!/bin/sh but
including bashisms is a problem. Would Ubuntu have a problem with
including bash? George: isn't bash POSIX compliant? Ted: yes, but
there's extra stuff, not supported on POSIX. Two problems: creating a
strict POSIX checker, clear and concise definition of the bash
extensions if we include bash. Mats: also, hash-bang isn't in the spec.
Darren: python and perl? Jeff: we require people to run the
interpreter manually, but in practice, people rely on hash-bang.
Darren: use of env? Jeff: env is in LSB. No specific path is recorded,
though. Might be a symlink in /usr/bin for distros putting it in /bin;
/usr/bin/env is nearly universal when hash-bang scripts use env. Mats:
env is not mentioned in FHS.
Jeff: reiterate: all LSB 4 tasks should be a bugzilla bug or a project
plan. Get those set up now, or be prepared to have things put off until
Mats: rename for ChipHopper apps should happen in next 10 minutes.
More information about the lsb-discuss