On 8 July 2015 at 20:01, Peter Crosthwaite <peter.crosthwa...@xilinx.com> wrote: > So you have caught about 10 broken PULLs in the last 48h :) I'm > guessing you have something heavily scripted, is the test system > available? Id like to run as much as I can over the multi-arch work in > total as that has widespread change.
My scripts are at https://git.linaro.org/people/peter.maydell/misc-scripts.git/tree (in particular pull-buildtest does the testing and remake-merge-builds tells you the configure runes for them). But I'm really not doing anything particularly exciting here: it's just running 'make && make check' for a bunch of different configs on an x86 box (plus doing it on an ARM 32-bit Linux box and on OSX). (And the linux-user tests from the wiki tarball. I ought to do something with the local tweaks I did to that makefile I guess.) Most of the things I find with the pull buildtests are just "this doesn't build in this compile environment". The main problem with the last 48h has been everybody assuming they can put a pull request in at the last moment. This is simply not physically possible given that it takes over an hour to process each one, and it's pretty risky given the chances of there turning out to be a "doesn't build" failure... thanks -- PMM
