In bug 910662, bustage is believe to have been caused by a commit made by a random bot, with no bug number listed: https://git.mozilla.org/?p=external/caf/platform/hardware/libhardware.git;a=commitdiff;h=34c1545405fad9fd42d1e7f8aa0b997e8d139604 I don't know why the bot makes those commits, who owns the bot, where it is syncing from, who made the original commit or what bug number the change was reviewed/discussed. This is just not scalable :-(

Note I don't even know if this repo is a mirror of a github repo, or it it is our own standalone repo (a whole other documentation problem/demonstration of how over-complex our B2G build story is right now). :-(

external/caf means codeaurora so we need to be watching those

Web view at: https://www.codeaurora.org/cgit/quic/lf Why a separate site?!

mwu might know more here.

We use codeaurora's fork of android for a number of devices. We have no control over them, though we can ask for fixes if there is bustage. Beyond that, the best we can do is freeze the commits used and update them on our own schedule.

Yeah, we definitely need to manually update which version we are using from external repositories. That way we can see when updating to a newer version breaks our internal tests and make sure to fix that bustage before we update to a newer version. How do we make that happen?

I suspect bug 899969 + the ability to mark certain repos as ones that the bot doesn't auto-update the manifest entry for, is the most practical way forwards here. (A human would then test + modify the pinned revision in the manifest).

Will be fixed by bug 910745