Do we need to be able to machine-edit the DEPS files even if we were to assume everything was pinned on the branch (and we had auto-rollers as needed)?

I'd like to start doing the latter ASAP.

Are you talking about the public branch DEPS? If we get that going as the branch source-of-truth, with the autorollers, then we don't need to do any edits to the branch buildspecs because there basically won't be anything left in the branch buildspecs. But I was assuming moving the source-of-truth was a little ways off still, and I kind of need the branch buildspecs cleaned up now to properly set the "checkout_src_internal" variable for recursedeps entries (for Issue 807318).

Yes, I was talking about the public branch DEPS.

I don't know why it would still need to be a little ways off, but I'd guess that's your call. I think that's a different way of me asking "what's keeping us from just fixing/switching to the public branch DEPS"?

(presumably pinning src-internal, at least, and possibly pinning/flipping ios and clank?)

Ideally the ios/clank (and other internal) stuff would be in src/DEPS, but that depends, at a minimum, on code search handling the "checkout_internal" conditions properly (and there might be other issues that haven't cropped up yet).

We could continue to have the buildspecs use the internal DEPS files, and only move the branch source-of-truth for the public stuff, but that could make the transition even more confusing than the current situation (i.e. "if the deps entry is in src/DEPS, it's updated by the branch autoroller, so don't touch it, but if it's in some_internal/DEPS, update it in the branch buildspec as before, until we eventually move those to src/DEPS, then don't touch it, ...").

Also, I think we want the branch CQ setup first, don't we? Or is that implied by having the autorollers working? Aren't there still branch checkout issues with bot_update or something?

I'm not sure if we actually need a CQ, or if the autoroller could just auto-commit things (we might want the former even if the latter is possible).

Let's try to schedule a meeting next week w/ all of us and hammer out a plan for getting this done. I'm a little worried that there are too many small tasks on too many different plates for this to be done in a timely manner otherwise.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/chrome/tools/release/scripts/+/bf6b3012e22a5098ec3e08ed8405b5e6e923840c

commit bf6b3012e22a5098ec3e08ed8405b5e6e923840c
Author: Michael Moss <mmoss@google.com>
Date: Fri Feb 09 19:50:34 2018

branch CQs today are not fully supported, there are quite a bit of caveats due to tech debt in CQ, cq.cfg, bot_update that I'd like to clean up, but there wasn't high priority to do so before. 

If you have more specific branch CQ requirements in mind, and i'll tell you what's missing in CQ for this.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/chrome/tools/release/scripts/+/9198cab96b5465415c59ef62ead941f38b3f35a4

commit 9198cab96b5465415c59ef62ead941f38b3f35a4
Author: Michael Moss <mmoss@google.com>
Date: Sat Feb 10 02:55:12 2018

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/chrome/tools/release/scripts/+/7a15bf5093059638602b0128c68b2a8a20198e4a

commit 7a15bf5093059638602b0128c68b2a8a20198e4a
Author: Michael Moss <mmoss@google.com>
Date: Tue Feb 13 06:18:01 2018

I'm going to close this, since it has already addressed the issue I was initially concerned with. I'll move branch CQ discussion to its own bug (or is there an appropriate bug for that already?).

per-branch CQ: issue 719954

sg, thanks