Hello,
An updated seed for Android platform is available as http://android-build.linaro.org/seed/uniseed.tar.gz . The requirement when preparing this seed was to make one which includes all repositories used by our manifest. Unfortunately, with such approach we just had swing of pendulum from one extreme to another. The old seed tarball was based only on (now pretty dated) pandaboard manifest, so other boards had to fetch increasingly more source with it. The new seed became not only comprehensive, but also very big and thus sluggish to download: its size 14Gb which is 4+ times increase over the previous one.
So, it's clear that we'll need other iteration(s) to find good compromise, but in the meantime the new seed fixes issues we had with older one: increasing rate git fetch failures and overall fetch times. With new seed, totals source preparation time is ~10mins. With the old one, we had ~9mins in the best cases. For the worst, I specifically did some statistic analysis and plotted graphs:
https://wiki.linaro.org/Platform/Android/SeededBuilds#Benchmarks_and_Statist...
As they show, there were builds were source preparation took almost half an hour. They also show that besides background slowly increasing checkout times, it big time increases happen in jumps (apparently when big repository, like kernel, being added/changed).
To use the new seed, REPO_SEED_URL should be set to the URL provided:
REPO_SEED_URL=http://android-build.linaro.org/seed/uniseed.tar.gz
I made a number of builds using new seed (~pfalcon/*-uniseed) which didn't show any issues (I don't expect any "what actually gets built" concerns - we've done with those during initial seed builds leveraging).
So, I would like to propose switching to the new seed once release is done, for example, this Saturday, to have results by Monday. Please let me know if this sounds ok.
Thanks, Paul
Linaro.org | Open source software for ARM SoCs Follow Linaro: http://www.facebook.com/pages/Linaro http://twitter.com/#%21/linaroorg - http://www.linaro.org/linaro-blog