1515106 - Add Raptor option to disable cache

Status: RESOLVED DUPLICATE of bug 1518975

(Testing :: Raptor, enhancement, P2)

Description

[Robert Wood [:rwood]]

As discussed on IRC it would be nice to be able to make it easier to run Raptor with the cache disabled (i.e. browser.cache.disk.enable = false).

It is possible to do that now by adding a 'preferences=' line in the Raptor test INI and pushing that to try.

However ideally it would be good to have a '--disable-cache' Raptor cmd line arg, that would work locally as well as add support for something like './mach try fuzzy --disable-cache' for Raptor tests.

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

Don't we have the cache disabled? I can see a kinda large impact on page load tests which get run in multiple iterations.

If successive page loads are all coming from the cache, we would always see shorter load times. Interesting is that for the mean value all iterations seem to be taken into account. The first load will always cause a larger variation, and more noisy results.

Browsertime seem to also not care about that:
https://www.sitespeed.io/documentation/sitespeed.io/best-practice/#how-many-runs-should-i-do-on-the-same-page-to-get-stable-metrics

Comment 2

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

Well, looks like browsertime clears the cache between each iteration:
https://www.sitespeed.io/documentation/sitespeed.io/best-practice/#clear-browser-cache-between-runs

So what would be the benefit of having such an option when the cache is cleared anyway?

Comment 3

[Dave Hunt [:davehunt] [he/him] ⌚BST]

This feels like a duplicate of bug 1518975, which added support for cold page loads. Whilst this doesn't disable the cache it does ensure that the cache is empty for each iteration.