Increasing prio, as it turns out lots of things are still broken, so we're blind to a bunch of potential issues. In particular, https://codereview.chromium.org/1396143002 messed up the client-side vs server-side detection, which broke (non-exhaustive list):
- NewTabPage.SuggestionsType
- NewTabPage.LoadTime.MostLikely

Once we have the ntp_tiles component, and we're also using it on the desktop NTP, all the UMA recording should happen there.

Fixit candidate :)

Given the amount of blindness this causes, this is definitely a high priority now. Once fixed we should even check out merge options to older branches.

IMO it's not really a regression, or at least not a recent one - the NTP UMA has been (slightly different variants of) broken for at least a year. The offending CL above landed 6 months ago, and things were already broken before that.

I agree it's a high priority, but IMO this needs a major restructuring - "point fixes" for the known-broken things obviously don't work. As such, it won't be a good candidate for merges.

Whatever we can do to fix the emission of data in M53, we should do it. Once we found a mitigation, we should decide if it's possible to merge it back to M52. It's really bad for us to be blind until September (M53 launch timeframe).

It's a regression, just one we caught very late and might have inherited.

Chris, thanks for taking a look!

I've been reading through the code and I feel like I'm starting to understand the various paths that are supposed to be used, but without a better understanding of what the breakage is, I can't identify the problem.

The difference in most_visited_single.js doesn't appear to be relevant, since the stats are (supposed to be?) updated through search_tab_helper.cc without going through the HTML/JS at all.

On the other hand, the "flush" behavior in NTPUserDataLogger::LogEvent() looks weird, and so does load_time_.

I think right now, the stats are "supposed to be" updated from both SearchTabHelper and most_visited_single.js, hence the most recent breakage.

Before Jan's recent CL, SearchTabHelper was supposed to do the logging, but it's not the right place (yet), because it doesn't know about MostLikely, only about TopSites.

There's still no actual description of the breakage here. No data? Partial data? Duplicate data? Garbage data?

SearchTabHelper doesn't know about MostLikely? Isn't that what's being tested here:
https://cs.chromium.org/chromium/src/chrome/browser/ui/search/search_tab_helper.cc?l=399
based on a value from from InstantService:
https://cs.chromium.org/chromium/src/chrome/browser/search/instant_service.cc?l=326
in a callback invoked by the suggestions service, which provides MostLikely data?

Breakage:
NewTabPage.LoadTime.MostLikely contains no data. NewTabPage.SuggestionsType always says "client" (same underlying problem).
There's quite possibly more breakage that I'm not aware of yet.

The problem is that the SuggestionsService (as used in InstantService) isn't actually active; it's behind a field trial that was never launched.

As far as the specific metrics in question, is http://crrev.com/390621 not sufficient to make them show up again? Is that something that could be merged to M52, or was it early enough (1.5mo ago) that that isn't necessary?

...maaybe? It looks like NewTabPage.LoadTime.MostLikely is being recorded again since M53, but the numbers don't quite add up - NewTabPage.LoadTime.MostLikely+NewTabPage.LoadTime.MostVisited have fewer samples than NewTabPage.SuggestionsType. Maybe there's a reason for that; I'm not sure when exactly all of these are recorded.

I suspect that some of the metrics might get recorded multiple times per NTP load due to the weird flushing behavior. It's also possible that everything is working as intended now, but that needs to be verified - right now I don't really trust any of these metrics.

You can see the state of all histograms in about:histograms - that should be good enough to sanity check things at least.

No longer blocked on other things, and not working on this.

Summary of the current state of affairs:
https://docs.google.com/spreadsheets/d/1W9eXR0G1GiJbd8SCX_mGSVqnxnTy-r7NsJKvGa0LO1c

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c7c5c817f22c6785a44a681bc562d05c6c4646d0

commit c7c5c817f22c6785a44a681bc562d05c6c4646d0
Author: sfiera <sfiera@chromium.org>
Date: Fri Jul 08 14:15:40 2016

Kill NewTabPage.NumberOfMouseOvers.

It's not clear what it would mean for the behavior to be correct, let
alone whether it actually is.

BUG= 611341 
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:closure_compilation

Review-Url: https://codereview.chromium.org/2091323002
Cr-Commit-Position: refs/heads/master@{#404362}

[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/resources/local_ntp/most_visited_single.js
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/resources/local_ntp/most_visited_util.js
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/ui/search/search_ipc_router_unittest.cc
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/ui/webui/metrics_handler.cc
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/ui/webui/metrics_handler.h
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/ui/webui/ntp/ntp_user_data_logger.cc
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/ui/webui/ntp/ntp_user_data_logger.h
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/browser/ui/webui/ntp/ntp_user_data_logger_unittest.cc
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/chrome/common/search/ntp_logging_events.h
[modify] https://crrev.com/c7c5c817f22c6785a44a681bc562d05c6c4646d0/tools/metrics/histograms/histograms.xml

Moving this nonessential bug to the next milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7668db462821887491792b20435d9bab1d749da1

commit 7668db462821887491792b20435d9bab1d749da1
Author: treib <treib@chromium.org>
Date: Wed Jul 13 16:59:51 2016

Add NewTabPage.TileType and .TileTypeClicked on Android

There already are versions of these split by tile source, but the accumulated totals weren't there.

BUG= 611341 

Review-Url: https://codereview.chromium.org/2131133003
Cr-Commit-Position: refs/heads/master@{#405199}

[modify] https://crrev.com/7668db462821887491792b20435d9bab1d749da1/components/ntp_tiles/most_visited_sites.cc