don't kill the OS incompatibility messenger.
1)
http://stackoverflow.com/questions/6062822/whats-wrong-with-strndup
stdndup() is POSIX 2008, but apparently not available on OSX and Windows
it could be made potentially application global (e.g. a local "stdndup.h")
2)
free() memory at pointer "current_dir", once we are done.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
O_TEXT is the default mode for fctrl's open() and on windows created
files, line endings are counted by fstat() as CR+LF adding an extra
byte for each line. the result from this is that, while the file still
can be read into a buffer, the read() return (ret) has a different
size compared to the previously allocated buffer, breaking at:
if (ret == mem->size)
a solution is to open() the file in O_BINARY mode, which should
technically suppress the EOL translation.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
[ Fixed to work under real operating systems that don't need this crap.
"Here's a nickel, kid, go and buy a real OS". - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The existing code set the filename to the full path of the last input file
and didn't set the path at all. Instead we now split the existing filename
into its path and file component and set up the choser accordingly.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When figuring out which cylinders to change in a multi-dive edit, we
already ignored the beginning and end pressures. But it turns out to make
more sense to also ignore the Nitrox / Helium settings.
Imagine you do a number of dives - for some reason your dive computer
records the wrong cylinder size in the downloaded logfile (like my uemis
does all the time). Dives will likely have different Nitrox percentage,
but you should still be able to simply fix the cylinder size for all dives
at once.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
It's an easy thing to do, but the for-loop ends up being pretty ugly, so
hide it behind the macro.
It would be even prettier with one of the (few) useful C99 features:
local for-loop variables. However, gcc needs special command line
options, and other compilers may not do it at all. So instead of doing
#define for_each_dive(_x) \
for (int _i = 0; ((_x) = get_dive(_i)) != NULL; _i++)
we require that the user declare the index iterator too, and the use
syntax becomes
for_each_dive(idx, dive) {
... use idx/dive here ...
}
And hey, maybe somebody actually will want to use the index, so maybe
that's not all bad.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The multi-dive case does fine, but the single-dive case (used when
adding a dive, for example) was somewhat confused between the dive index
(which is the location in the dive array) and the dive number.
Fix this by just passing the dive pointer instead (where NULL means to
use the current dive selection).
Reported-by: Jacco van Koll <jacco.van.koll@gmail.com>
Root-caused-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ok, so the widget doesn't actually *do* anything, but this is where you
would add dive printing settings for things like "print list" vs "print
profiles" etc.
Printing just a dense dive table (no profiles etc) is being discussed on
the list, maybe starting the scaffolding will inspire somebody to do
something about it ...
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now that the last commit tried to avoid changing the child selections if
the selected group partially matched, we should always [un]select all
children when we actually decide to change something.
Before, it would try to minimize selection damage by stopping
[un]selecting when it hit a child that already matched the selection,
but since we minimize damage differently, the all-or-nothing approach is
better, and gets us sane behavior when the group is collapsed.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This tries to avoid the problem mentioned in commit972669d6363c ("Rework
dive selection logic"), where a selection of dives hidden by collapsing
the group gets forgotten about by gtk. It does so by always marking the
group selected when it is collapsed with any selected children.
We also avoid selecting new children when a group is selected that
already has at least *some* children selected already. This way we do
minimal damage to existing selections when working with dive group
selections.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This completely changes how we keep track of selected dives: instead of
having an array listing the selection ("selectiontracker") or trusting
the gtk selection information, just save the information about whether a
dive is selected in the dive itself.
That makes it trivial to keep track of the state of selection across
group collapse/expand events, or when changing the tree view model. It
also ends up simplifying the code and logic in other ways.
HOWEVER, it does currently (re-)introduce an annoying oddity with gtk:
if you collapse a dive trip that has individual selections, gtk will
forget those selections ("out of sight, out of mind"), and when you do
*new* selections, the old hidden ones remain.
So there's some games required to make gtk do sane things. We may need
to either explicitly drop selections when collapsing trips, or make sure
the group entry gets selected when collapsing a group that has
selections in it. Or something.
There may be other issues introduced by this too.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This adds the ability to expand/collapse all the dive groupings in the
divelist from the divelist right-click context menu.
Should we perhaps add it to the top 'Dive' menu too?
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Previously when only one dive was selected, the Stats notebook page would show
the statistics for all dive. That creates a very illogical behavior when
clicking on the different dive groups in the dive list. The stats page would
always show how many dives where in a group when the group was selected, except
when there was only one dive in the group, in which case the statistics for all
the dives were shown.
With this change we also show the statistics for the selected dives, even if it
is just one. If you want the statistics for all dives, simply select them all
(Ctrl-A or Command-A on a Mac).
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Linus change in commit bcb9f67819bc ("Make the notebook portion (dive
notes/equipment/info) a scrollable window") created a really ugly default where
the notebook Dive Notes always ended up with a vertical scrollbar. This picks a
much saner default layout for the panes.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This makes things start up with the wrong size, which is somewhat
annoying, but by doing so avoids a bigger annoyance, namely that the
three panes move around when moving between dives.
In particular, if the initial dive didn't have much of an equipment
list, the initial size allocated for the notebook is fairly small and
determined mainly by the size of the the Dive Notes page. However, when
you then scroll around in the dive list, you might hit a dive with lots
of equipment, and suddenly the panes dividing the different parts of the
subsurface application window will jump around to make room.
That's horribly annoying, and actually makes things like double-clicking
dives in the dive list not work right, because the first click will
select it, and cause the dive to move around (so the second click will
hit a totally different dive).
Now, making the notebook be in a scrollable window means that if the
size of the notebook changes, it might get a scrollbar, but the panes
themselves do not move around.
The initial sizing of that thing being wrong is annoying, though. We
need to figure out a separate solution to that.
[ Side note: currently it uses GTK_POLICY_NEVER for the horizontal
scroll-bar, just to avoid the horizontal size also starting out wrong,
which is *really* nasty. If we can solve the initial size issue, we
should make the horizontal scroll-bar be GTK_POLICY_AUTOMATIC too. ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We now pick one hour after the end of the currently selected dive as the
default starting time for the new dive to be added. If multiple dives
(or no dives) are selected, we default to current time as before.
The "one hour after the end" is just a random (but not unreasonable)
assumption for the surface time if you add multiple dives.
Suggested-by: Miika Turkia <miika.turkia@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This shows the number of dives in the grup in the divelist header field,
and also picks the location from the first dive that *had* a location,
so that if any dive in the group has a valid location, the group will
have a location.
It also makes double-clicking a dive group expand/collapse that group.
Requested-by: Miika Turkia <miika.turkia@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the suit and weightsystem support of Subsurface when importing
divelogs from JDiveLog. (They were previously included in the notes
field as support for these fields was missing from Subsurface.)
After import the weightsystem is undefined and weight unit is the
default of Subsurface. Unfortunately the weight field in JDiveLog is
text field and might contain pounds and kilograms mixed in seemingly
random order. Thus 2 pounds of weight might be transformed to 2 kg.
Signed-off-by: Miika Turkia <miika.turkia@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The "filename" variable was only initialized when the user accepted the
name, so cancelling the file save would randomly use an uninitialized
pointer.
Reported-by: Miika Turkia <miika.turkia@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The code iterates over a list that can be NULL, but happily dereferenced
it anyway. Oops.
This function really should be split up and commented more.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This fixes the case of the dive duration being zero, or being shorter
than the assumed ascent/descent time.
Reported-by: Lutz Vieweg <lvml@5t9.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I lied in the commit message for commit 0468535524a3 ("When editing multiple
files, don't override existing equipment entries"); the changes there did
not parallel the logic for the string entries. Now I think it does.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
I missed one instance where a callback function needed to be passed the widget
index w_idx in the signal_connect function. It got passed a pointer to the
model instead which of course blew up when trying to dereference the array with
that "index".
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We have removed a menu separator from the gtk gui and that was still
referenced in the Macos code. And just in case, we are now testing
for the widget for the other separator to be non-NULL before
destroying it.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Pull dive selection fixes from Dirk Hohndel.
This hopefully fixes the common cases. Dirk is cursing gtk. We may
need some gtk selection guru to explain things.
* 'fixes' of git://git.hohndel.org/subsurface:
Another selection fix
More fiddling with the selection
The corner cases are getting more and more artificial. Without this patch,
the following can happen:
Select one or more dives in an (expanded) dive trip. Now collapse that
trip with the little triangle. Select a different trip. The previously
selected dive(s) are still part of the selection (as you can see, for
example, in the statistics tab).
With this patch the scenario above works as intended (all the dives in the
new trip are selected), but we have another corner case:
Just as before, select one or more dives in an expanded dive trip.
Collapse that trip and ctrl-click on another trip. Now you lose the
originally selected dives.
Frankly, if you ctrl-click to add more dives to your selection - just
don't collapse the trips the dives are in?
As this new corner case seems even more artificial than the previous one,
I consider this patch an improvement. But fundamentally I am just battling
all the ways in which gtk's selection handling is messed up. When I get
the selection call back I cannot tell if this is a new selection or an
incremental selection (i.e., a shift-click or ctrl-click).
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
As expected, this is pretty subtle to get right. But with this change the
code becomes simpler and more straight forward, I think. If the dives in a
group are collapsed, we don't even try to make gtk keep track of their
selection status - we explicitly do so ourselves. This avoids the
artificial expand / collapse around our attempt to force gtk to allow us
to select children that are hidden. But if a dive is expanded, then we
trust gtk to get things right.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Pull miscellaneous fixes, mostly UI stuff from Mikko Rasa.
Both this and the pull from Pierre-Yves Chibon created a "Save As" menu
entry and logic. As a result, there were a fair number of conflicts,
but I tried to make the end result somewhat reasonable. I might have
missed some semantic conflict, though.
Series-acked-by: Henrik Brautaset Aronsen <subsurface@henrik.synth.no>
* 'misc-fixes' of git://github.com/DataBeaver/subsurface:
Add a separate "Save as" entry to the menu
Changes to menu icons
Improved depth info for dives without samples
Divide the panes evenly in view_three
Pull patches to change behavior on exit from Pierre-Yves Chibon.
Pierry-Yves explains:
"When someone opens a file, change something in it and try to quit, the
program asked whether the data should be saved.
If 'Ok' then it shows the save-window and the user can choose to save
the file or rename it.
My habits in such case would be that since I opened a specific file, I
want to save to that specific file, therefore, when I press 'Ok', I
want it to save automatically to the file I opened.
So I have been working on changes that do:
- When a file has been opened by the user, save to this same file if
the user is 'Ok' while closing.
- Add a 'Cancel' option to the pop-up window that offers to save the
file while closing.
- Add a 'Save As' entry in the file menu."
* 'change_quit2' of http://ambre.pingoured.fr/cgit/subsurface:
Add a 'Save As' entry in the menu.
Allow to cancel while trying to quit and the data was changed.
When the file has been opened rely on it to save.
Linus' code dropped the const qualifier from the start rating. While
fixing this I stared some more at get_combo_box_entry_text and realized
that the existing code could potentially change the "old" pointer and then
pass it to free(). Tsk-tsk-tsk.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Commit 2f773b97e0 ("multi-dive editing: don't change already set data
for other dives") didn't get the multi-dive editing quite right: even if
one of the dives in the list of changed dives has an empty field, we
should *not* fill it with the edit data unless that edit data was
actually changed.
So compare the new data with the original master data, and if they
match, do nothing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add a "Save As" entry in the "File" menu allowing the user to specify the file in which to save
the data. This is useful as we no longer offer this option through the "Save" entry while the data
had been opened from an existing file.
Signed-off-by: Pierre-Yves Chibon
So far, when trying to quit while the data was changed the offer
was "Save" or "Don't save". Now, you can also "Cancel" which will
bring you back to the main window.
This allows you to re-save the data in another file.
Signed-off-by: Pierre-Yves Chibon <pingou@pingoured.fr>
When a file is opened, we keep it in memory and when you try to
quit while the data has been changed, propose to save back to
this same file.
Signed-off-by: Pierre-Yves Chibon <pingou@pingoured.fr>
Instead of having "Show Temp", "Show Cyl", etc in the Preferences dialog,
rename the group as "Show Columns" and remove "Show " from all the
checkboxes. The dialog is tighter/nicer this way.
Signed-off-by: Henrik Brautaset Aronsen <subsurface@henrik.synth.no>
Pull selection tracking fixes from Dirk Hohndel:
"I just gave up on gtk tracking our selection. Way too much pain. The
implementation below has seen some testing with the debugging code
enabled and seems to work - but it needs more banging onto it, I'm
sure.
Ideally I'd like to leave the debug code in, ask people on the mailing
list to play with it and report any inconsistencies. After that I'll
be happy to remove it again."
* 'tree2' of git://git.hohndel.org/subsurface:
Stop relying on gtk to track which dives are selected
When editing multiple dives at the same time, don't change fields that
have already been set for a dive, unless the old field contents match
the currently selected ("master") dive.
So when you edit multiple dives, you can set the dive master or buddy
(or suit etc) for all of them in one go, but if one of them already has
that field set, it won't be modified just because you set the other
ones.
Acked-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We spend way too much effort trying to get gtk to manage the dives that
are selected. The straw that broke the camel's back is that gtk forces us
to expand any nodes that we want to select - so selecting a summary entry
for a dive trip forced us to expand all the dives in the dive trip. Which
as Linus pointed out really sucked from a user experience.
So instead we now completeley ignore gtk's weird idea of what is selected
and what isn't and simply track things ourselves. We still need to play
some games with gtk to make sure that the correct rows are SHOWN as
selected, but still, the overall code seems much cleaner.
This commit contains a bunch of debugging code that is ifdef'ed out -
this is extremely useful to make sure I didn't mess anything up, but
eventually I'll want to remove that again as it just looks ugly in the
code.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This fixes the bug that triggered the SIGSEGV that Linus worked around
earlier. I had forgotten to update this call path to the
edit_multi_dive_info function.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The multi-dive editing is broken if you right-click on the dive
text-fields (instead of the divelist). This just avoids the SIGSEGV, it
doesn't really fix the editing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull dive-trip grouping from Dirk Hohndel:
"This turned into an updated pull request for the tree2 branch where I
implemented the date based grouping - but is actually a very different
topic: this adds the ability to edit multiple dives (and fixes some
issues with the dive editing overall). The reason for that is that it
reuses some of the infrastructure that I implemented in the tree2
branch for tracking the selected dives. More details in the commit
messages."
* 'tree2' of git://git.hohndel.org/subsurface:
Switch from date based to dive trip based grouping
Redo dive editing
Fix selecting and unselecting summary items
Apply sort functions to the correct model, don't select summary entries
Maintain selected rows when switching between list model and tree model
Create duplicate list model so sorting by columns works again
Improve tree model implementation
Allow date based grouping
Linus HATED the date based grouping - too much wasted space visually
("three levels of grouping are way too much") and asked for dive trip
based grouping instead.
This is a quick change to do just that, with an assumption that no
dive in 3 days means it's a new trip.
This also changes the summary entry to display a location for the trip,
for now we pick the location of the (chronologically) first dive of the
trip.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This commit addresses two issues:
We now can add / edit / delete equipment from the edit dive dialog
We now can edit multiple dives at once
The latter feature has some interesting design constraints:
It picks the 'selected_dive' as the one to start the edit from - so if
this dive already has some information filled in, that information needs
to be overwritten before it is stored in all of the dives. Similarly, only
changes to the cylinders or weightsystems are recorded. Also, the notes
field is not editable in the multi dive edit mode (as that didn't seem
useful).
The workflow seems to work best if using the multi-edit right after
importing new dives from a dive computer. The user then can select all the
new dives and only needs to edit things like location, divemaster, buddy,
weights, etc. once.
This commit will create some obvious conflicts with the commit that adds
exposure protection tracking. It was implemented on top of the tree_view
changes as it reuses some of the infrastructure for tracking the selected
dives.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
For simplicity and shortness, throughout subsurface exposure protection is
simply referred to as "suit".
Add the fields to the data structures, add the column to the dive_list
and the preferences dialog (once again with it being turned invisible by
default). Support loading and saving of the suit information.
Display the suit information in the Dive Info pane (this may be a bit
controversial as people could argue this should be in the Equipment pane)
and allow editing of the suit info, with our usual support for completion
and drop down lists to pick from.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
In fill_one_dive(), cylinder and location strings are obtained via
get_string(), which needs to allocated a litte bit of memory.
After passing the two pointers ('cylinder' and 'location') as arguments
to gtk_list_store_set() it is safe to release them.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
The dive list now seems to behave intuitively.
In order to do this we had to intercept the select function in addition to
having a selection-changed callback. That way we can simulate the
multi-level selection and unselection that was missing.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We only set up the column specific sort functions for the default (tree)
model, which caused us to not sort correctly in the list model.
This commit also somewhat cleans up the handling of selecting summary
lines in the tree model, which includes the very first selection made at
program start (which happens to be the very last dive).
But it still doesn't work the way I expect it to work (i.e., the correct
row is not highlighted). Fundamentally I would prefer clicks on the
summary lines to instead select (or as ctrl-click, possibly deselect) all
the dives under that summary entry. Still TODO.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
