1568911 - 2kb of cache returns after deleting from about:preferences#privacy -> Cookies -> Clear data...

Status: NEW

(Firefox :: Settings UI, defect, P3)

Description

[pciapa]

Attachment: 2jDDl.gif

Steps To Reproduce

1. about:preferences#privacy
2. Cookies and Site Data
3. Clear Data... button
4. check-uncheck any checkbox (needed for make popup active)
5. CMD + a (select all text)
6. Button Clear
7. again CMD + a (select all text)
8. blue button 'Clear now'

Result: cache is not 0, but can be even more than before pressing 'Clear now'

Spotted on OSX and WIN10
Bug is intermittent, happens 7 of 10 runs

Firefox version - 69 0b7

Comment 1

[Jared Wein [:jaws] (please needinfo? me)]

Thanks for the bug report. Can you please run mozregression to find when this started happening?

https://mozilla.github.io/mozregression/

Comment 2

[pciapa]

(In reply to Jared Wein [:jaws] (please needinfo? me) from comment #1)

Thanks for the bug report. Can you please run mozregression to find when this started happening?

https://mozilla.github.io/mozregression/

Sorry, at the moment running mozregression is not in my scope of tasks, and I don't have personal time.

I will find out how to help you, my team has not yet established workflow on creating new bugs, sorry.

Comment 3

[pciapa]

On win7 and osx found that by default there is not 0 bytes,
while on Linux it is 0 bytes.

Comment 4

[pciapa]

Attachment: not 0 bytes.png

not 0 bytes

Comment 5

[Jared Wein [:jaws] (please needinfo? me)]

I can reproduce this. Setting needinfo to Honza as he worked on bug 915296.

Comment 6

[Honza Bambas (:mayhemer)]

(In reply to pciapa from comment #0)

Result: cache is not 0, but can be even more than before pressing 'Clear now'

Where exactly do you see it? It's not clear from the screenshot at comment 4.

Note that the browser may simply make a background request that will fill the cache just after you have cleared it.

Likely INVALID.

Comment 7

[pciapa]

(In reply to Honza Bambas (:mayhemer) from comment #6)

(In reply to pciapa from comment #0)

Result: cache is not 0, but can be even more than before pressing 'Clear now'

Where exactly do you see it? It's not clear from the screenshot at comment 4.

Note that the browser may simply make a background request that will fill the cache just after you have cleared it.

Likely INVALID.

On screen 4 there is 2.0 Kb instead of 0 bytes

Also note that this 2.0 kB I have on clear profile on OSx machine while running same profile on Windows machine shows 0 bytes

Comment 8

[Honza Bambas (:mayhemer)]

If it's 2jDDl.gif then the UI is wrong from some reason and this bug sits in the right component.

Note that firefox is often doing background requests, specifically after start with a new profule, so the cache is at 0 only for a very short time. You can examine the content with about:cache.

Comment 9

[pciapa]

Attachment: w7.png

2.0 KBs here while it needs to have 0 bytes. On osx and linux it do have 0 bytes, but on Windows 2.0kbs

Comment 10

[Mike Conley (:mconley) (:⚙️)]

Hi pciapa,

It sounds like mayhemer suspects that a background network request is repopulating your cache after you've cleared it.

One way to test that hypothesis is to repeat your experiment, but to make sure that there's no network connection before clearing the cache. This can be as simple as disconnecting from the wired / wireless network before running the experiment.

If the network connection is down, are you still able to reproduce?

Comment 11

[:Gijs (he/him)]

I suspect this is a dupe of bug 1518256.

Comment 12

[Mike Conley (:mconley) (:⚙️)]

(In reply to :Gijs (he/him) from comment #11)

I suspect this is a dupe of bug 1518256.

What's the best way to confirm? Should we have the reporter provide some about:cache information after clearing the cache?

Comment 13

[:Gijs (he/him)]

(In reply to Mike Conley (:mconley) (:⚙️) (Catching up from PTO) from comment #12)

(In reply to :Gijs (he/him) from comment #11)

I suspect this is a dupe of bug 1518256.

What's the best way to confirm? Should we have the reporter provide some about:cache information after clearing the cache?

Probably. I mean, we could also just fix the code to lie. Just make it say 0 bytes. It'll be more than 0 bytes if/when you refresh; that seems OK.

Comment 14

[Honza Bambas (:mayhemer)]

Yeah, and somebody will one day file a bug that "there is data in the cache folder, but UI says it's 0 - fix it!!". We decided to reflect what we actually allocate on disk rather than lie. Cache is a living thing, so it constantly changes.

Can we close this bug as WONTFIX now or is there some real issue left?

Comment 15

[:Gijs (he/him)]

(In reply to Honza Bambas (:mayhemer) from comment #14)

Cache is a living thing, so it constantly changes.

Sure, I think everyone on this bug is aware of that.

Yeah, and somebody will one day file a bug that "there is data in the cache folder, but UI says it's 0 - fix it!!". We decided to reflect what we actually allocate on disk rather than lie.

I think this is good, in general. But from a layperson's perspective, the current UI is suboptimal. If I tell software "get rid of a thing", and the result the software shows me is "tada, the thing is still here", that's super confusing. Yes, everyone on this bug understands why (there are background requests!) but most users won't - it just won't make sense to them, and we do nothing to explain it.

We could replace the message that includes the size with "The cache was cleared." after clearing it. We could even set a timeout to update it again [some interval] later with a then-up-to-date number, or to add a button (again, after some interval) that replaces the "it was cleared" message with one indicating the current size. Perhaps UX could come up with an even better option here.

Either way, the point I'm trying to make is, although the UI is technically correct at the moment, it's confusing for users. We should inspire confidence, not confusion.

Comment 16

[pciapa]

(In reply to Mike Conley (:mconley) (:⚙️) (Catching up from PTO) from comment #10)

If the network connection is down, are you still able to reproduce?

Without network connection I can't reproduce this.

Cache of local resources that I load without connection is deleted to 0 bytes.

ps. sorry for late answer, I was on vacation, could not to notify everyone here.

Comment 17

[Honza Bambas (:mayhemer)]

(In reply to :Gijs (he/him) from comment #15)

(In reply to Honza Bambas (:mayhemer) from comment #14)
Perhaps UX could come up with an even better option here.

Yes, because I really am not the right person to decide on this. It's tricky. This is not the first bug around this issue. Note that animations (anything that changes in front of users' eyes) were rejected (can't find the bug#)

Comment 18

[Honza Bambas (:mayhemer)]

Emanuela, can you please help find someone to decide from the UX perspective what the behavior here should be?

The point is that we want to clearly communicate to the user that the content cache has been cleared completely after the button press. What we do now is to clear the cache and then refresh the size in the UI. The size returned by the HTTP cache API returns a non-zero, because there are indexing structures and possibly background requests that fill the cache quickly again. We don't want to lie about how much disk space we take for the cache, but want to make sure people understand the cache (aka history) was deleted.

Thanks.

Comment 19

[emanuela [ux]]

My recommendation would be to work with the content strategy team to be sure our message is on point here.

Comment 20

[:Gijs (he/him)]

(In reply to emanuela [ux] from comment #19)

My recommendation would be to work with the content strategy team to be sure our message is on point here.

OK.

(In reply to Honza Bambas (:mayhemer) from comment #18)

Emanuela, can you please help find someone to decide from the UX perspective what the behavior here should be?

The point is that we want to clearly communicate to the user that the content cache has been cleared completely after the button press. What we do now is to clear the cache and then refresh the size in the UI. The size returned by the HTTP cache API returns a non-zero, because there are indexing structures and possibly background requests that fill the cache quickly again. We don't want to lie about how much disk space we take for the cache, but want to make sure people understand the cache (aka history) was deleted.

Thanks.

Michelle, can you or Betsy provide some suggestions about what we should do here?

Comment 21

[BugBot [:suhaib / :marco/ :calixte]]

Clear a needinfo that is pending on an inactive user.

Inactive users most likely will not respond; if the missing information is essential and cannot be collected another way, the bug maybe should be closed as INCOMPLETE.

For more information, please visit auto_nag documentation.