Get rid of chroot inside this script, and not mounting mod_for_test_scripts into rootfs any more.
The reason caused the failure here is once you chroot into the mounted rootfs directory, all binaries were searched inside the rootfs. But all binaries inside the arm rootfs was for arm hardware, so not executable on the host linux machine(intel).
Review URL: http://codereview.chromium.org/845007
This change may be controversial -- it will basically make the x86-generic
full buildbot fail if tests fail to build.
There are a few options:
1. do this and see how it goes;
2. don't do this and rely on the test failing to run;
3. do this but also change buildbot to pass --nobuildcheck to build_autotest
Review URL: http://codereview.chromium.org/848005
When grabbing buildbot image and test artifacts, properly install the test artifacts as if they were built by the user (with the user as owner).
Review URL: http://codereview.chromium.org/703001
2. I dont fully understand why we need sudo inside this script. So I am removing it. Since I could not enter sudo password inside crontab. Please let me know if I missed anything and I could revert.
Review URL: http://codereview.chromium.org/666002
This saves almost 1 minute (out of 6) on my build host for the current full
build_autotest.sh. The savings will be more significant once the Chrome tests
(200MB+) also get into autotest.
Alternatively, we could still build the binary package and use it as part of
archive_build.sh (i.e., just copy it there instead of creating a new one).
But most people don't run archive_build so it may be better to not create
the binary package as part of build_autotest.sh.
Review URL: http://codereview.chromium.org/661473
sync_build_test runs build_autotest that way.
archive_build bzips up the autotest artifacts.
run_remote_tests can run autotest artifacts from a prepackaged directory. Note that because build_platform is in a different repository than these other files, it's unfortunately forced to be reviewed separately in http://codereview.chromium.org/661197 and committed independently.
Review URL: http://codereview.chromium.org/660189
For example, ./build_packages --<tab> will try to complete all supported
./build_packages flags (or print them if completion is non-unique).
This completion could be used outside chroot.
Review URL: http://codereview.chromium.org/661096
If you source this script inside chroot, you would get <tab> completion for the --board= argument for some build scripts.
For example,
. bash_completion
./build_image --board=x<tab>
may complete to
./build_image --board=x86-generic
if x86-generic is the only available board starting with "x". If not, it will list available boards with the given prefix.
We could extend this to work with setup_board and other scripts. Also, it may be nice to extend it to work outside chroot.
Does it make sense to make this part of the regular dev chroot by default? If so, not sure what the best way to do that. Maybe install this as a hard-host-depends and source it from the chroot user .bashrc?
Review URL: http://codereview.chromium.org/652129
Remove make_local_repo stage as it's not used by either build process.
Make portage build the default (--noportage still builds the old way).
Also add total time to build. A run of sync_build_test without args
on my machine takes 2:20min. Defaulting to --jobs=1 due to
instability reports.
Review URL: http://codereview.chromium.org/646015
ecee13e13945c7df79e397012054206cac6cb2f0 back to the previous revision.
This new file should be identical to the one at
ce18a0349a88ffbe01a2e88526aeb271b54dec60
Review URL: http://codereview.chromium.org/646057
