D15507: [MountPointObserverCache] Update remote and slow mounts less frequently

[Harald Sitter]

sitter added a comment.


  As mentioned on IRC I was thinking that we should simply bump the overall poll interval to 1 minute instead of splitting timers. The additional code seems hardly worthwhile considering the original 10s value is really odd if you do some quick maths... from IRC:
  
    [13:18] <sitter> I am not sure how those 10 seconds came to be really ... it seems to me the diff is not noticable with anything but tiny disks. I mean, say you have 300mb/s throughput, which seems generous even on the same partition, you'd be having 3gb in 10s. what's a common SSD size? 320? So you'd be looking at a 1% change from one poll to the next. on my system the bar seems to be some 75px, so within one poll frame the bar wouldn't even move. and that's already 
    [13:18] <sitter> assuming I'd be logging around such huge amounts of data
    [13:20] <sitter> at the same time if you consider the entire thing for an HDD it's an even weirder default
  
  ... and then you also have to ask yourself how often would a user have to move such substantial amount of data around that it'd justify such a short poll interval. After all, 99% of a partitions live data <3gb/s are being added/removed, so the general situation is next to no change with the occasional massive change.

INLINE COMMENTS

> anthonyfieroni wrote in mountpointobservercache.cpp:58
> Maybe true will be better.

Doesn't really matter, does it?

If we can get a KMountPoint we'll defer the decision there.
If the protocol is clearly not local it gets mark slow.

Changing the default value doesn't remove the need for either of those condition.

But, I think we can simplify this by initializing the value correctly, i.e. move the scheme check up:

  bool isSlow = (KProtocolInfo::protocolClass(cachedObserverUrl.scheme()) != QLatin1String(":local"));
  [......]
              isSlow = mountPoint->probablySlow();

> broulik wrote in mountpointobservercache.cpp:86-87
> Why? There are protocols that aren't `file` but still local (e.g. `desktop`)

Yeah, I don't see what's wrong with this?

REPOSITORY
  R318 Dolphin

REVISION DETAIL
  https://phabricator.kde.org/D15507

To: broulik, #dolphin, elvisangelaccio, sitter
Cc: anthonyfieroni, kfm-devel, feverfew, spoorun, navarromorales, firef, andrebarros, emmanuelp
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20180917/df5e9371/attachment.htm>