This makes the git object save logic also check out the changes in the
working tree and index if the branch we save to is checked out. It used
to be that we would just update the object store (and the branch ref, of
course), but leave any checked-out state untouched.
Note that if the working directory is dirty (ie you have made changes by
hand and not committed them), the checkout will skip any dirty files and
report it as a warning to the user. However, the save still succeeds
(since the _real_ save goes to the backing store).
NOTE NOTE NOTE! Both loading and saving very fundamentally work on the
git object store level, and if you are working with a checked-out branch
and make modifications to the working tree, saving will not touch those
dirty files (so that you can try to recover your edits manually in the
working tree), but it's worth pointing out that subsufrace loading state
will totally ignore the working tree.
So the only way to make subsurface *see* your changes is to commit them.
Having edited state checked out in the working tree will only confuse
you when subsurface first ignores it on reading, and then refuses to
touch the checked-out state on writing.
Put another way: working with a checked-out branch is now _possible_,
but you need to be aware of the limitations.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When some arguments like --help and --version are passed to the
executable, we don't need to create the UI at all.
This patch separates the QApplication creation which is at first
only needed to parse the arguments and then if exit() is not
called from subsurfacestartup.c, we can call some of the
"init" methods such as setup_system_prefs(), fill_profile_color()
etc.
At this point init_ui() can be called which no longer needs
to accept the command line argument list.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Qt5 uses different widths for some of the poly. lines
in the profile. Setting an explicit value fixes that.
Tested-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When using Up/Down arrows to scroll the tag list it always selected the
first item in the list and doesn't scroll.
Fixes#468
Signed-off-by: Gehad elrobey <gehadelrobey@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
After editing values in cylinders or weights hitting save will not save
the changes to save the edit one must move the focus to a different
field first this is fixed by losing the focus before saving the changes
Fixes#412
Signed-off-by: Gehad elrobey <gehadelrobey@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When building and running subsurface out of tree, marbledata was
inaccessible to subsurface.
Signed-off-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
display.h: dc_number is unsigned int, thus a couple of warnings
may pop-out.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
After the dive log has been closed, the ProfileWidget2 does not clear
the pn2 po2 and ph2 gas line.
This patch cleared the three lines after dive log closed.
Acked-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Hugle <wulong@comp.nus.edu.sg>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
There are no utf8 in those strings, and we can translate them as
everything else with tr() instead.
QApplication::UnicodeUTF8-part is deprecated and removed in Qt5.
Signed-off-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The text for these two sections has been corrected and updated
in order to be ready for the release of V4.1. The very first
paragraph of the manual is expanded substantially.
Signed-off-by: Willem Ferguson <willemferguson@zoology.up.ac.za>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The selection logic was a bit random: some places would return NULL if
the dive computer index was out of range, others would return the
primary dive computer, and actually moving between dive computers would
just blindly increment and decrement the number.
This always selects the primary computer if the index is out of bounds,
and makes sure we stay in bound when switching beteen dive computers
(but switching between dives can then turn an in-bound number into an
out-of-bounds one)
Fixes#464
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
I had intended to amend the previous commit with this before pushing it
out. This changes the comment that pointed out the bug that the previous
commit fixed.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Turns out we unconditionally set all events as visible when redrawing the
dive - even with a comment that this should take into account if the event
is visible. Oops.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
A new figure is included to reflect the actions of the new profile buttons
on the calculated ceilings.
A number of small edits on sections 4 and 5 of the user manual are done.
These sections are now finalised for the new release.
Signed-off-by: Willem Ferguson <willemferguson@zoology.up.ac.za>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This got lost when we switched to the new profile.
Remove event works. Hide events does call hide() on the DiveEventItem but
for some reason it stays visible. I'll hope for one of the more
experienced Qt people to fix that.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
If the user tries to delete a cylinder that is in use and that action is
rejected, the cylinder widget (and the whole dive) should not be put into
edit mode. After all, nothing changed.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Add simple test dive to show the erroneous pp graph in case of
multiple periods of high p02. See commit aa0cd792bb for the fix.
Signed-off-by: Jan Mulder <jlmulder@planet.nl>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We always checked if the tank was in use by the planner / dive editor -
even if we were not in dive edit mode.
With this patch, when not in dive edit mode, we check our cylinder "used"
flag instead.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
commit d681bdcb63 ("Only use default cylinder for first one") has a
stupid bug in that it only calls get_tanksize for the first tank. That's
of course completely bogus.
Thanks to Linus for catching this.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When downloading from a divecomputer it makes little sense to assume that
all tanks are the default cylinder.
There's a good case to be made for having a default first cylinder (you
always dive with your own cylinder, or you are always on a dive boat with
AL80 tanks), but in multi-cylinder situations this is much more likely to
cause unintended harm; for example for those dive computers that always
report their maximum number of cylinders, even if some of them aren't
used. Here setting a default cylinder turns those entries from obviously
empty into something that appears to have meaning (i.e., cylinder type is
filled in) even though this was just a default added by Subsurface.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We used to fall back to an AL80 default cylinder, but that meant that a
user who doesn't want a default cylinder at all had no way to indicate
that.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Especially in O2 decompression parts of a dive, the pp02 is typically very
close to the threshold value (normally 1.60 bar). The old implementation
of the pp profile graphs assumes that there is exacty 1 consecutive set of
samples that needs to be in the "warning color". This results in an
erroneous display of the mentioned graphs, connecting multiple episodes of
too high pp with bogus lines in between.
This fix generalizes the pp graph logic to allow for multiple segments of
high pp, each to been drawn seperately in the "warning color".
Signed-off-by: Jan Mulder <jlmulder@planet.nl>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This reverts commit e70bbb637e
Linus' solution in commit 27c36ec4cf ("Improved handling of git syntax
names with no git repository") is much better and makes my hack
unnecessary.
This makes "is_git_repository()" return non-NULL for all file names that
match the git name pattern, even if we don't find an actual git
repository there. That way, we won't fall back to writing out an XML
file with an odd filename.
If there is no actual git repository, we return a special invalid dummy
pointer, and then the git reading and writing routines will catch it and
return the appropriate error.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Before this change when the user gave a git style filename, say
/tmp/mydives[myname], if Subsurface couldn't create a repository with that
branch, it instead saved an XML file to this exact name, with '[' and ']'.
That clearly is not the desired behavior, so report an error instead.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The last figures in the manual using the Unity desktop have been replaced
with equivalent figures using Gnome 3. There should now be no figures
remaining from Unity.
This is a revision of the profile part of the manual in the light of the
changes for V4.1. Text has been moved around, some new text added or
rewritten. Graphics has been replaced to reflect the UI for V4.1,
including Tomaz's button bar and Louisa's new buttons for that bar.
Signed-off-by: Willem Ferguson <willemferguson@zoology.up.ac.za>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The "report_error()" interface is a lot simpler, although some of the
C++ code uses QStrings which make them a bit annoying, especially for
the varargs model. Still, even with the explicit conversion to UTF8 and
"char *", the report_error() model is much nicer.
This also just makes refreshDisplay() do the error reporting in the UI
automatically, so a number of error paths don't even have to worry. And
the multi-line model of error reporting means that it all automatically
does the right thing, and reports errors for each file rather than just
for the last file that failed to open.
So this removes closer to a hundred lines of cruft, while being a
simpler interface and doing better error reporting.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This makes the error string just be an internal "membuffer", which the
GUI can fetch and show when errors occur. The error string keeps
accumulating until somebody retrieves it with "get_error_string()".
This should make any write errors actually show up to the user.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This at least avoids marking the dive list as unchanged on a failed
write, and propagates the error further up the stack.
We still don't show the error string in the GUI, though. I'll start
doing that next, I think.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Note! This just returns the error (and uses "report_error()" to generate
a string that is currently printed to stderr). Nothing actually *uses*
that error return yet, and we don't show the error string in the GUI.
Baby steps.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Commit 13e2210d75 ("Allow remote branch names when reading a git
object tree") made it possible to read (but not write) remote branches,
which is very convenient when you just want to look at somebody elses
dives in a shared repository.
However, it was really quite stupidly done - both overly complicated,
and overly restrictive.
It's much better and simpler to just allow general git revision
specifications, which includes branches (both remote and local) as a
simple case, but also allows general git revision expressions. So you
can tag things, and use a tag-name instead. Or you can say that you
want to look at the previous save, by using the "branchname^" syntax.
Or, you can use the git reflog, and do things like
subsurface ~/scuba/[linus@{two.days.ago}]
to see the dives that your repository contained two days ago.
Obviously, you will not be able to save to this kind of ref-spec (and I
really will have to make error handling work better), but for browsing
state it's quite useful.
And in git terms, this is actually simpler than the "lets try to first
see if we have a local branch of that name, and then if we have a remote
one", as shown by the fact that this removes more lines than it adds.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This makes subsurface remember the git source commit of the dive data.
If you save to an existing branch, subsurface will now complain and
refuse to save if you try to save if the existing branch is not related
to the original source. That would destroy the history of the dive
data, which in turn would make it impossible to do sane merging of the
data.
If you save to a new branch, it will see if the previous parent commit
is known in the repository you are saving to, and will save parenthood
information if so. Otherwise it will save it as a new parentless commit
("root commit" in git parlance).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
By simply storing the coordinates based on the scene (instead of trying to
map them to real coordinates) the overlay position is correctly restored.
Also remove the redundant positioning before readPos is called.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is the quick hack to read from a remote branch, which allows you to
look at other peoples branches when sharing a git tree.
Note that the "remote" part of "remote branch" is the _git_ meaning of a
remote branch: it is the local cached copy from a remote. This does not
imply any kind of network traffic - but if you have done a "git fetch"
to get branches from some other source, you can now use the remote
branch-name to see them in subsurface.
Also notice that you should *NOT* save the end result. It will "work",
but it won't do what you think it does. Saving does not update the
remote branch, it would create a new *local* branch with that same
branch-name, and since it's a new branch, it would do so with no
parenthood information. So you'll be very very confused.
I think I'll add code to remember the parent when loading from a git
repository, and then use that remembered information when saving. So
then you could create a real local branch with real history. But that's
an independent issue from this loading case.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Instead, just encode the git repository information in the filename.
We want to make it much harder to make it match a real filename, but to
still allow easy browsing with the file manager interface. So the git
repository "filename" format is the path to the git repository
directory, with the branch name encoded as "[branch]" at the end rather
than the "path:branch" format that we used in the descriptor file.
[ For example, on Windows, a filename like "c:\my.xml" could be
interpreted as the branchame "\my.xml" in the repository in the
directory "c" ]
In particular, with this model, no filename that ends with ".xml" could
possibly ever be considered a git repository name, since the last
character of a git pathname is always ']'.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is a bit tricky because we are using a plain widget for
a window and don't have a class for it (req. more source files).
Also for the table model to update we need to create a new
YearlyStatisticsModel instance each time. At least, in that regard
we can re-create the model each time refreshDisplay() is called.
This patch adds a couple of private variables that are used
to manage the memory of the yearly statistics model and window
and also close that same window on MainWindow::closeEvent().
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The topmost Z order of items in the profile should be:
- background (poster / logo)
- toolTipItem
- rulerItem
...
This mostly fixes the ruler being under other elements.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
A couple of problems with the ruler:
- the rotated text doesn't look very well at all and interpolation doesn't
help it much
- measuring towards the right most part of the profile makes the text go
out of the screen
To solve these issues and attempt to improve the ruler this patch does the
following:
- place the text at the bottom of the lowest of the start and end points.
this way the line will never intersect with the text
- clamp the x position, so that the text doesn't ever leave the screen
horizontally
- place a white background behind the text so that it will cover text and
graphics under the ruler item
(TODO: place the ruler on top of everything else)
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Only the change to subsurface.pro is authored by me.
The rest is generated (subsurface_source.ts) or created by the translators
on Transifex.
This adds Greek, Hungarian (partial), Latvian (partial), Romanian, and
Turkish (partial) translations.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
For most dialogs it would be much safer to make them Qt::WindowModal to
the parent (MainWindow).
For now we are not doing this for the preferences dialog as there are
situations where the user might want to be able to move it around and even
interact with the main window.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
If the newly created save tree is identical to the parent commit tree,
don't bother creating a new commit. We are already fully up-to-date.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
I didn't think that one through: the version string is already saved in
the commit message, and so saving it in the tree object is redundant.
Now a little redundancy doesn't hurt, but having the tree object depend
on th esubsurface version _does_ end up being annoying: it means that as
you update the subsurface version, doing a data save will result in a
different tree SHA1 even if none of the data changed.
Which doesn't actually matter right now, since we always create a new
commit anyway, but my plan was to skip the commit creation if nothing
changed in the tree. And saving the version string defeats that if you
are a subsurface developer and the subsurface version keeps changing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We don't actually much use the trip list any more, and it's possible we
should simply get rid of it. I hadn't added the trips to the trip list
when loading them, and everything worked fine.
Well, *almost* everything worked fine.
There is one use of the list of trips, and that's the "clear the trip
index for each trip before saving them". That literally seems to be the
only non-debug use of this list, but when we didn't add the trips to the
list, the trip index never got cleared before saving trips.
And even that is unnoticeable for the *first* save event, because the
trip index will have been clear before that.
But on the *second* save event, if the trip index doesn't get cleared
before saving, the saving code will look at the index, say "Hey, I
already saved this" and skip the trip.
So if you loaded the trips from a git repository, and then saved things,
everything worked fine. But it you saved things a *second* time,
nothing would get saved at all, because all the trips were marked as
saved already.
Anyway, I think the real solution is to get rid of the pointless trip
list, and just use "for_each_dive()" to find all the trips, since that
list clearly is just more pain than gain. But in the meantime, this
makes the git loading add the trips properly to the list.
Signed-off-by: Linus "oops" Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
