Android build 784

[Dirk Hohndel]

On Sun, Feb 07, 2016 at 10:09:12AM +0100, Jan Mulder wrote:
> Tested the latest build. That is, with the "edit mode stays around when
> exiting it via the Android back button" commit reverted. So again, an own
> build.

Please remind me why you aren't testing the APK that I'm uploading?

> On 06-02-16 22:40, Dirk Hohndel wrote:
> >Unfortunately it seems that a "fix" that I attempted for the "edit mode
> >stays around when exiting it via the Android back button" may have a) not
> >fixes that in a satisfying way and b) have introduced crashes (at least
> >for Jan).
> Some more info on the crash I see. It seems very similar to the "endless
> loop implicit size" issue we have seen before, and which appeared to have
> gone away by Ricks 2nd attempt from the DiveEdit header layout formatting.
> So, its is probably not related to the mobilecomponents change. Not that
> that really seems to matter because the Plasma team is considering a
> different fix.

My guess is that there is still something broken in the way we build the
size of the layout and it randomly triggers.

> Some more on performance. For the people that are on IRC, they know the I
> (re)found a function that is pretty heavy on computation;
> get_pr_interpolate_data. Used to get proper gas consumption line in those
> cases that the divecomputer does not transmit actual pressure data.
> Obviously, this is very dependent on divetime, but in addition, it is very
> dependent on sample interval. For example, on my phone, it takes approx 10
> seconds to render a 4 hr dive registered with a OSCT3 with sample interval
> of 2 seconds. I was very surprised that an older 6 hr dive rendered in under
> 1 second. This one is, however, registered with a Reefnet Sensus Ultra
> datalogger set to a sample interval of 10 seconds (7000 samples on OSTC3,
> 2200 samples on Sensus). All this is definitely worth some further
> investigation, as it scales almost quadratical.

I have just pushed a commit that doesn't run these calculations at all if
you have no pressure data at all. That doesn't address the quadratic
runtime of the algorithm, but it does simply remove the delay if there's
nothing there to interpolate, period.

/D