On Sat, Aug 11, 2012 at 3:47 PM, Jon Chiappetta
<Jon.Chiappetta(a)senecacollege.ca> wrote:
Thanks to Peter for pointing out that the previous times I had were
overall
task times and not actual buildarch times so I re-wrote my script to gather
that data instead. Basically the script just adds/sums up the time it took
for each arch so (i686 + x86_64) and (armv5 + armv7) and compares/totals
them up.
Can I suggest you compare the two and take the longest build. The
different arches are built in parallel so adding them both together
isn't really a good reflection of the build time. Also I suggest
discarding anything built by a guru or smarttop as again they're not a
reflection of what the ultimate HW will be for building.
If indeed my numbers are correct this time, it looks like we took
an extra 32 days or about a months worth to complete? Keep in mind
these
times also include our slowest builders, where in which if you assumed a
build farm composed of only our biggest/best machines, it'd probably be a
lot faster (Brendan was recommending I filter out the slow ones). You can
verify the task times by clicking on the links in the page below:
Yep, hence the suggestion to discard the other ones and maybe take the
other build time for the said package.
The above said it's looking good and we're getting there. We're
certainly not an extra month behind mainline.
Peter