V10K Concerns, what happens at Chrome OS build 10000? |
|||||
Issue descriptionChrome OS has been around for a while, and every day we generate ~3 new version numbers off of ToT. As we have achieved a build number of over 9000, we should start to consider what happens when we roll over to a 4th digit. For example I suspect we have scripts floating around that assume the build is 4 digits in pattern matching, that might start to act weird. I am not sure of the best way to audit this, or whom would be best to do so.
,
Nov 18 2016
+ge folks too Great point, although I do not recall any incidents when going from 999. to 1000. back in M15! Many things have changed since then though so it is good to check/validate
,
Nov 18 2016
GoldenEye pads the version components with zeroes, so we don't need to worry until we get to 1e7, which is almost 10 millennia away :) Given that current versions are not sortable by string already (because of the components: 1234.5 < 1234.10) I don't expect this to be an issue AFAICT.
,
Nov 18 2016
Glad to hear Goldeneye will be ok with it. I suspect it will mostly work in the Chrome OS code base as well, I am not sure where but I recall seeing (or writing) at least one place where identifying a string as looking like a Chrome OS version number involved a 4 digit regex though. We just need to keep on the look out for such cases.
,
Nov 18 2016
We didn't plumb in anything expecting only 4 digits on the GE side, so agreed with Moises that we don't expect this to be a GE issue.
,
Nov 18 2016
We are roughly 100 days, or 3 months away.
,
Nov 18 2016
i don't recall writing or seeing any 4 digit assumptions in general in the code base can't we patch the chromeos_version.sh file in the chromiumos-overlay to make it 10k+ and then run a remote trybot ?
,
Nov 18 2016
That is a good point, lets see what happens in https://uberchromegw.corp.google.com/i/chromiumos.tryserver/builders/release/builds/7237 which has https://chromium-review.googlesource.com/#/c/413042/
,
Nov 18 2016
Try the same build with "--hwtest" as well, so we can exercise the lab a bit.
,
Nov 20 2016
The trybots seem to be happy, the version "10002" does show up in the log, and it looks right. The trybot still looks a bit different since we don't get to exercise the logic that uprevs the version, or uses the same manifest version sync stage, but this is promising.
,
Dec 26 2017
,
Jan 2 2018
[It appears that a bunch of old cros issues bulk-added the "Infra" component recently, but they should probably be "Infra>Client>ChromeOS".]
,
Mar 30 2018
Ha ha ha. We ended up having a postmortem outage due to this. |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by bhthompson@google.com
, Nov 18 2016