On 06/30/2013 04:59 AM, Daniel Veillard wrote:
I have been watching the automatic mails since they were
(thanks I now have yet another in memory counter, I was actually
mentally happy when it crossed 4000 thinking we're getting closer to
50%). What is the limitating factor ? I don't think there is a too
big depth of package dependencies, so if the main factor isn't some
lack of computing power, I would have assumed that unlocking some
key package (e.g. python, gtk libs, ...) would allow some quick
progresses, but it smells like it's progressing purely lineary. Is
that dependant on converting each of the package (like auto* updates
when they lack aarch64 support ?)
Cole (I assume it's mostly you based on mails and commits samples
I saw), would you share the story there ?
Not sure who Cole is.
Progress is currently limited by number of builders. There are about
300 source rpms in the build queue at present and more being added as
deps get built. As we get deeper into the queue the packages remaining
tend to have a wider array of dependencies and take longer to build.
Remember all these builds are being done by aarch64 simulation, so
installing and linking with a wider array of libraries has a material
impact on build time.
There are still some packages that once built will unlock a large number
of additional builds. The single biggest blocker today is openjdk.
We're working on bringing it in. Unlocking ghc (waiting on llvm which
is waiting on gcc 4.8.1) and erlang would also unlock many.
There are a number of packages that have been attempted, but failed to
build. Everybody is welcome to grab one of these and fix it- don't let
us ARM team members claim all the honour! The list of broken packages
Quick start guide for contributors is here:
Brendan Conoboy / Red Hat, Inc. / blc(a)redhat.com