New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 740083 link

Starred by 1 user

Issue metadata

Status: Untriaged
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

Plan to remove obsolete histograms and flags

Project Member Reported by dgn@chromium.org, Jul 7 2017

Issue description

We currently keep obsolete histograms, enums, flags etc around so that they can still be used to display details for older chrome versions (IIRC). The issue is that the histogram files are painful to diff/blame and review because of their sheer size. 

For some items we don't even mark them obsolete. I remove some chrome://flag entries but can't remove the associated xml items for example (https://chromium-review.googlesource.com/c/561719/)


Could we move obsolete items to a separate xml file? It could make the size more manageable and the current state of options easier to identify.



 
Cc: holte@chromium.org
Steve was indeed planning to split off obsolete histograms into a separate file. 
  For the flags entries, we don't have any concrete plans currently.  Ideally, we'd keep an append-only mapping somewhere server-side, and drop the requirement to keep histograms.xml up-to-date with historical information.

FWIW, both git and gerrit seem to be fine in terms of diffing the files.  Which tools are you having trouble with?

Comment 2 by dgn@chromium.org, Jul 10 2017

Good to know it's already been considered, thanks!

I remember rietveld giving up when trying to do a diff between intermediate patch sets. And it also just takes ages to load on codesearch. Or doing a blame there, etc.
The other idea we've discussed is having the server-side infrastructure store history, so that things could just be deleted at TOT and the dashboards would still know the historical descriptions and which things are obsolete (by virtue of them having been deleted.)

If we do that, then we don't need separate files for obsolete entries, so on the Chromium side it will simplify things a lot.
Re #2: Thanks! Rietveld did indeed have an issue with this file, which Gerrit has mostly fixed.  Codesearch is indeed still slow :(

Re #3: True, though some histograms provide some useful info in their <obsolete> message, such as what histogram replaced them.  It's admittedly not too common, though.

Sign in to add a comment