My recent builds are failing on repo init.
See: https://android-build.linaro.org/jenkins/job/linaro-android_panda-ics-gcc47-tilt-tracking-blob/126/console
+ repo init -u git://android.git.linaro.org/platform/manifest.git -b linaro_android_4.0.4 -m tracking-panda.xml --repo-url=git://android.git.linaro.org/tools/repo.git --reference=/mnt/jenkins/workspace/linaro-android_panda-ics-gcc47-tilt-tracking-blob/build/uniseed
/usr/local/bin/repo: line 1: syntax error near unexpected token `<'
/usr/local/bin/repo: line 1: `<html><head><title>Gitweb down temporarily</title></head><body>gitweb is currently unavailable until we finish importing jellybean</body></html>'
++ infrastructure_error
++ echo 'Caught infrastructure error - finishing build with '\''Not Built'\'' status'
Caught infrastructure error - finishing build with 'Not Built' status
++ exit 123
Ok, so what happened here is: we're getting "repo" initial binary from our servers, not google's (because they can be down any time). To get it, we use wget and download it from gitweb, because git doesn't allow to check out a single file from a repo (and because original google instructions also use wget, so we chaneg just URL, nothing else). Now, that gitweb was put down and replaced with a static page, we get that static page and trying to access, getting the error above.
More interesting question is though why this occurred just yesterday, while gitweb was disabled last week. It turned out that "linaro- android_ panda-ics- gcc47-tilt- tracking- blob" job had it's "labels" field in Jenkins config set to wrong value, which caused old build slave type to selected for it, uncovering the issue as described above. All other jobs use custom AMI build slave, which has correct "repo" preinstalled and thus not susceptible to gitweb being down.
linaro- android_ panda-ics- gcc47-tilt- tracking- blob was fixed, and it was checked that no other job has that issue.
What's left is probably to remove old non-custom AMI build slave types, because their presence leads to confusion now.