Lots and lots and lots of header files were being included without being
needed. This attempts to clean some of that crud up.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This cache give us a huge gain in performance, going from
17% moving the mouse frenetically to 9%, wich is quite acceptable.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We used to create a new QGraphicsRectItem everytime a Pixmap
changed. Since I'm pretty sure I deleted every bit of the
PictureItem before setting a new one, no leak was due, but this
version is safer.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We were calling this even if we didn't really change anything
and paths are expensive to paint.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We were recreating the PathItems (one for the outline, other for the real
text) for every call to setText. This was a very un-smart move.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Do the switching based on the index rather than the string (which is
translatable!).
Update set-points when turning on/off CCR mode (remeber: the rebreather
mode is stored in two places: the divemode and implicitly in the setpoitns.
Signed-off-by: Robert C. Helling <helling@atdotde.de>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Update O2 metabolsim rate and adopt default gas switch depths to
pSCR oxygen drop.
Signed-off-by: Robert C. Helling <helling@atdotde.de>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This reduces a lot of CPU time and makes the overall use of the tooltip a
breeze.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The tooltip animation had a fixed animation speed, this patch
honors the anim_speed on the preferences, and also disables
the animation completely if the speed == 0.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
There are a few calculations that go on boundingRect that can be avoided
if we simply store the result.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Correct pen and brush set. the ToolTip now is correctly rounded,
translucent and happy.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The rectangle is now correct, but the collors are still
wrong. I'm tracking that down - most probably I've set
the wrong pen or brush ( or both ) somewhere.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Those items were used to fake the background of the path item
but since the rectangle can be painted with a border and a
fill, this is uneeded.
The rect is still ugly.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
a rectangle is *much* faster to paint than a simple ShapeItem,
so this is a safer choice. We still need to create the paint
method so we can use the correct roundness for the rectangle.
Currently it's white with a 1px solid line - terrible. :)
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We were deleting / recreating the graphics background item for *every*
mouse movement. Now we are just creating the painter path; no more
allocations / desalocations, adding, removing from the scene. This should
make things a tiny bit faster.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
While analizing the code for the mouse movement I've discovered that
we did a lot of uneeded things: Set the color, the pen, the size
of a fixed-colored line, twice.
We also deleted-newed the same Pixmap / Text for every mouse movement
so now we reuse the 'entryToolTip' that consists of a huge line and
a pixmap, and after that we add the other tooltips that are not static
Also, reduced a lot the number of calls to expand() (that did a lot of
math).
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Strangelly, this method was being called even if the rectangle was the
same, so we deleted everything and recreated everything again. tsc tsc.
Some more improvement is needed but we are getting there.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Very often the rectangle of the ToolTip doesn't need to change but we were
calling and firing an animation for it for *every* mouse movement, even
when we didn't really needed it.
Now it will only fire something if the rectangles are indeed different.
From my tests we reduced the number of calls to the animatior by about 20%
using a real divelog as test.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is an attempt to make fewer calls to alloc functions when the mouse
is moving.
We were creating a membuffer, filling it (malloc / realloc), then freeing
it just after use. but we could simply hold that allocated area and reuse
it again.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We did three cals to mapToScene / mapFromScene on the mouse moveEvent at
the ProfileWidget2 where we only needed to call one in the common case and
two in the worst case.
This doesn't really help in terms of speed (unless you have a really old
cpu) but since it's code that gets called *very* often, it seemed a
reasonable thing to do.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The QPainter and the QPixmap were being created but never freed. A QPixmap
and a QPainter don't need to be created by new, they can be safely created
on the stack.
So, create them on the stack, pass them via const-reference
and use them correctly.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
After looking with great care at the result of the mouse movement
on the profile, and also playing a bit with callgrind I've found
out that one thing that we were doing wrong was the way we looked at the
items in the scene, by calling scene()->items with
Qt::ItemIntersectsShape, our shapes are very complex curves
with thousends of points and we have lots of them. and it usually
doesn't matter because *most* of the time we are getting the
tooltip information from 'get_plot_details_new', so no accessing
to items was necessary.
By changing the access from Qt::ItemIntersectsShape to
Qt::IntersectsItemBoundingRect we had a speedup of almost 500x in a
section of code that's very important, and the good thing, nothing bad
happened because one of the only things that we are using this code is to
get information from the events, not the curves.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Take instantMeanDepthLine out of the code. We have the moving average line
plus the exact data in the information overlay.
Signed-off-by: Cristine Guadelupe <cristineguadelupe@me.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Ignore QWebView instances in the preferences dialog when compiling under
Android, as QWebView is not yet supported under Android.
Signed-off-by: Joseph W. Joshua <joejoshw@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
I could have sworn we did this at some point. The vendors are already
sorted, but the products for each vendor should be sorted, too.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
After a partial download (because the user hits cancel or because there
was an error, if the user hits Retry the list of dives downloaded so far
should be cleared because we will simply try to re-download the same dives
again.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This will crash if for some reason current_dive is invalid.
And in general, when displaying information, we want to use the correct dc
in the displayed_dive, not the current_dc, which references the
current_dive...
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When changing to a CCR dive, add a setpoint change to the default setpoint
at the beginning of the dive. Otherwise add an explicit setpoint change to 0
.
Signed-off-by: Robert C. Helling <helling@atdotde.de>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is bigger and more invasive then I wanted, but it's hard to break it
down into smaller pieces. Here's what it does:
The former "Download" button becomes the "Download", "Cancel download" and
"Retry" button. So this button controls your interaction with the dive
computer.
The other two buttons are now purely "OK" and "Cancel" for the dialog.
"Cancel" discards what happened (much easier now that we download into a
different table), and "OK" adds the dives that were selected in our
selection UI (by default all downloaded dives) to the real dive_table.
And while redoing all this, I also redid some of the state machine
underlying the dialog. The biggest change that the user will see is that
partial downloads (after canceling or after an error) will still offer the
dives that were completely downloaded up to that point in the selection
menu.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We pass a different table to libdivecomputer (and the uemis code) and have
that table filled. And then we simply copy the dives from that table into
the real dive_table when the user accepts the download.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This will allow us to download the dives from the dive computer into a
separate table just for that purpose and not into the main dive_table.
I really dislike the code that's in place that dates back to the very
earliest code written for Subsurface. Dumping the dives straight into the
main dive_table seems really stupid to me.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Since we now track all clicks on the row, having the setData() actually
neutralized the action that we took on the clicked() signal.
Now you can select / deselect a dive, regardless where in the row you
click.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Yet another bug because the indices are inclusive. We need to start off
with the last being smaller than first and we need to adjust the row
count. It might be easier to just fix thing to make last be exclusive...
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This fixes the issue where there was no visual feedback when clicking on
the second or third column in the grid. It would actually change the
checked state of the checkmark internally (and you would see the new state
once you clicked on another dive), but it wouldn't give immediate visual
feedback.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The array we allocated was one entry too small.
On the flip side, let's just make sure we cannot call this with a negative
range. That would be bad, too.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Correct minor malfunction with CCR setpoint display. It was showing even
when the po2 display was turned off. This patch ensures that the setpoint
graph only shows when the po2 toolbar button is activated (and in addition
the appropriate checkbox in the Preferences).
Signed-off-by: willem ferguson <willemferguson@zoology.up.ac.za>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Previous code suggested a "upgrade" if your firmware where other than
whats parsed from the autofirmware call.
This switches that logic into only suggesting firmware upgrade if the
found firmware is newer than the firmware your device runs.
The previous logic was very annoying if you have a device running a
development version of the firmware. If you have taken that step of the
guided path, you shouldn't be told to "upgrade" to something older
every time you download from your computer.
Signed-off-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
