[lsb-discuss] LSB conf call notes for 2008-06-11
jeff at licquia.org
Wed Jun 11 09:03:13 PDT 2008
Attendees: Jeff Licquia, Ron Hale-Evans, George Kraft, Dan Kohn, Mats
Wichmann, Carlos Duclos, Russ Herrold, Stew Benedict, Vladimir Rubanov,
Alexey Khoroshilov, Robert Schweikert, Alan Clark, Ted Tso.
3.2 refresh beta 2.
3.2 SI. Stew: s390 problem, rebuilding now. Jeff: beta time. Alexey:
bug in SI: no fonts. Stew: should this be a showstopper for the SI?
Ted: maybe a new feature? Alexey: some apps won't run in the absence of
fonts. Ted: problem with taking time on 3.2. Jeff: maybe provide a
font package as an add-on for the SI, to prevent rebuilding? Seems like
the consensus. Stew will look into doing that; won't hold up the beta
waiting for it.
4.0 symbol list from Mats. Jeff: worth discussing these, so we can make
decisions and act.
ptrace. Ted: why do people use it in a non-debugging environment?
Robert: language implementors. Ted: problem is that ptrace is different
per-arch; lots of work. Mats: Google performance toolkit (which seems
to come up a lot). Sun has been trying to kill ptrace. Ted: ptrace
like ioctl, painful to specify. Jeff: original bug filed from wine.
Ted: wine/CrossOver isn't going to be LSB-compiant anytime soon. Mats:
google perf tools uses it to clean up threads on a signal. Ted: could
we pull someone in? Jeff will ask Dan Kegel for someone to discuss
ptrace. Ted: also discuss syscall, gettid, possible other "evil"
dl_iterate_phdr. Lots of apps use it: avg, f-secure, vmware, java,
google earth, google desktop, groupwise, adobe reader. Mats: not used
in google performance tools currently (claimed bug). Jeff: noted that
it's Linux-specific; any other objections? Mats: std. objection: need
an accurate spec. Ted: stable? Jeff: seems to be, latest change dates
back to glibc 2.2.5. Add? Mats not sure he knows the whole story.
Ted: odd, but seems useful.
dlinfo. No docs? Ted: concern is the tie to the dynamic linker. Jeff:
only request is internal; no apps use it. Maybe skip? Mats: makes sense.
Xext*. Mats: used by a lot of apps. Not documented, require a
structure that's really ugly. Ted: can we wrte a wrapper that doesn't
expose the structure. Mats: most extension libs have a query function;
this is something else. Java uses it for their self-tests. Ted: maybe
we tell the X people "write a new ABI, or we'll standardize the regular
one". Also, so many people are using it that it's a de-facto standard
already. Robert agrees. Ted: add as deprecated? Jeff will send a note
to the X folks regarding this issue. Robert: cross-post to lsb-discuss.
strverscmp. Used by a few apps. Ted: low cost.
alphasort/scandir. Discrepancy between POSIX/history and the glibc
spec. Need a test of some kind to see if glibc can take the more
precise version. Ted: check with glibc tip as well.
More information about the lsb-discuss