Other info:

  1. I can really see there is a branch [1] in the third_party/kernel.
  2. Sheriff and I can "repo sync" coral factory branch successfully and locally.

[1] https://chromium.googlesource.com/chromiumos/third_party/kernel/+/factory-coral-10122.B-next/chromeos-experimental

Assign owner to infra deputy first for visibility.

more and more factory / firmware branches met the same error including fizz, glados and others. And would expect to get support for upcoming factory build of partners. Thanks.

fizz: https://uberchromegw.corp.google.com/i/chromeos.branch/builders/fizz%20pre-flight%20firmware-fizz-10139.B/builds/51

master-release release-R64-10176.B is also in the same issue.

https://uberchromegw.corp.google.com/i/chromeos_release/builders/master-release%20release-R64-10176.B

Similar to  bug #801211 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/manifest/+/9269bd1ac03fef1c5b622e5aa2b1d283c4245f27

commit 9269bd1ac03fef1c5b622e5aa2b1d283c4245f27
Author: Bernie Thompson <bhthompson@google.com>
Date: Thu Jan 11 17:47:10 2018

Remove kernel-experimental branch from R64 manifest

BUG= chromium:801062 
TEST=None

Change-Id: I473611f74240a6f48b6e76d7a641d250698129c4
Reviewed-on: https://chromium-review.googlesource.com/862102
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: YH Lin <yueherngl@chromium.org>
Commit-Queue: Bernie Thompson <bhthompson@chromium.org>
Tested-by: Bernie Thompson <bhthompson@chromium.org>

[modify] https://crrev.com/9269bd1ac03fef1c5b622e5aa2b1d283c4245f27/full.xml

Wonder if some entity or person tried to create a branch named "factory-coral-10122.B-next" (and release-R64-10176.B-next and so on). That would mess up git quite nicely.

Also, there might be a problem if there was an infrastructure change which uses '/' as delimiter to parse branch names.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/chromeos/manifest-internal/+/5a30975300b1584044b144bf7be1f327e887d245

commit 5a30975300b1584044b144bf7be1f327e887d245
Author: Douglas Anderson <dianders@chromium.org>
Date: Thu Jan 11 18:29:08 2018

+jimhoff@ GoB git oncall

So removing the kernel-experimental branch did not help, it just fails on the next one.

I can check this out locally and it works ok, though I do start seeing it trying to use git gc on a cleanly wiped checkout which makes me nervous that our kernel repo has grown beyond some boundary that may be making us more likely to see corruption?

I think we are nearing the end of our expertise here, adding GoB git oncall for thoughts on why we might be failing checkouts intermittently like this. How would one normally debug these sort of 'Transient error string identified in STDERR: "fatal: Couldn't find remote ref refs/heads/factory-coral-10122.B-next/chromeos-experimental'?

This doesn't sound like a production emergency affecting many customers
(what oncall typically handles), but I am having trouble understanding the
details of the bug.
If it a production emergency, let me know. If it's not, go/gob-ticket is a
good way to get support. Also see our docs
<https://g3doc.corp.google.com/company/teams/gerritcodereview/support.md?cl=head>
on support in general.

Adding the chrome trooper; "[W git.go:283]" messages come from a
chrome infra tool.

The stuff that's failing is mediated by code in Chrome Infra, not
Cros Infra.

So, roping in a trooper.

I don't think 801322 is related.

After much tribulation, we've found and filed a GoB bug:  b/71872654

It ain't happenin' no more, and GoB team claims its fixed, so, let's
declare victory.