This is the exact same case as the previous commit, just for the writing
side.
Once again, it's the subsurface rfcomm iostream code that can return
DC_STATUS_SUCCESS with a byte count of zero when something goes wrong
with the write.
And once again, our libdivecomputer iostream code didn't try to be
robust and protect itself from that case.
The fix is equivalent, although slightly simpler, since the write side
doesn't have the whole timeout issue.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We had two independent bugs here, both of which needed to fire for this
to cause a problem. This fixes both of them.
The first bug was that our rfcomm code would return DC_STATUS_SUCCESS
with a zero-sized read when a timeout happened, or when the rfcomm
socket had disconnected. That makes absolutely no sense. We should
return DC_STATUS_TIMEOUT on timeout, and DC_STATUS_IO if the socket has
disconnected without any data.
The fix to this is to make the whole rfcomm iostream read logic much
simpler: there's no need to loop at all for partial results, because the
libdivecomputer iostream side will do the loop for us (and handle
partial results much better: it knows if the target backend can handle
those partial results or not).
The second bug was in our libdivecomputer iostream read() function,
which reacted very badly to this bad return value. This updates our
libdivecomputer branch to one that is more careful about things.
Reported-by: linuxcrash <albin@mrty.ch>
Debugged-by: Jef Driesen <jef@libdivecomputer.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The diveListNotifier.divesSelected() signal is used to inform the
models of a selection change. It sent the current dive as a second
parameter. This is redundant, because the only sender of the signal
sets current_dive just before sending the signal. Remove the
parameter, which appears to be an artifact.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Grammar-nazi ran
git grep -l 'indexes' | xargs sed -i '' -e 's/indexes/indices/g'
to prevent future wincing when reading the source code.
Unfortunatly, Qt itself is infected as in
QModelIndexList QItemSelection::indexes() const
Signed-off-by: Robert C. Helling <helling@atdotde.de>
Since I learned while trying to implement this that getting sub-second
resolution time in portable C99 is hard (especially for someone who is
used to the comfort of std::chrono and Howard Hinnants date library) the
timer-implemetation from libdivecomputer is now copied to the subsurface
source.
Signed-off-by: Christof Arnosti <charno@charno.ch>
Thanks to the new USB serial implementation also that complex special-casing
is no longer needed. This should do the right thing now.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Create a C string (which the caller needs to free) with the executed commands
in this session.
The detour via the callback allows us to not make the corelib depend on the
commands, which is nice for tests, export-html, and smtk2ssrf.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
In the QML code we pass ids around. I had assumed that there already was a reverse
lookup function, but I wasn't able to find it. So I added it.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Use the undo-command for importing dives also on mobile. This should make the
whole disconnect-model shenigans unnecessary.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
The goal is to send the signal for the correct dives / divesites
and thus not having to reload the whole model.
Right now the mobile UI does not yet catch the diveSiteChanged signals.
[Dirk Hohndel: small fix to ensure that we trigger a save to storage]
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This finishes the spliting of the GPS fix application:
One function for collecting the fixes, one for application.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Not that this would make any noticeable difference, but out of
principle, let's use Qt's string-literal macro for string-literals.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Make the application of the GPS fixes in two runs: first
collect dives and fixes, then apply the fixes. This will
simplify turning the application of GPS fixes into an
undo-command.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
We know that we cannot support native USB, USB HID, IRDA, and USB
storage on Android.
On the flip side, don't try to force the long broken FTDI download.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Since the Android USB stack and subsequently the usb-serial-for-android
driver have problems with read-timeouts, the read-timeout is now
implemented in AndroidSerial.java. Also, DC_STATUS_TIMEOUT is returned
if there are less bytes returned than expected.
Different chipsets seem to behave differently with
usb-serial-for-android. On CP210x the read blocks until there is some
data here, but on FTDI the chip seems to return whatever is currently in
the buffer (so 0 bytes if the buffer is empty). This different behaviour
should be mitigated by the changes by this commit.
Signed-off-by: Christof Arnosti <charno@charno.ch>
Implement the libdivecomputer API in Java and create C/JNI translation
layer.
[Dirk Hohndel: whitespace harmonization - yes, some of this is Java,
this still makes it much easier to read for me;
also changed the FTDI conditional compilation to make
sure we can still use that for mobile-on-desktop if
necessary]
Signed-off-by: Christof Arnosti <charno@charno.ch>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is a quick hack to reduce the noise in the log file when chasing other
bugs. Maybe this should not be enabled on release builds, but right now I don't
think the harm that having this in would do.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
For reasons which I don't yet understand, when plotting a dive
whose first cylinder is not cylinder 0 and then plotting a dive
with only one cylinder, it can happen that for the latter
explicit_first_cylinder() returns an erroneous value.
This is due to the way in which we copy the dive to be plotted
to displayed_dive.
For now, make sure that no invalid cylinder is returned to avoid
crashes. This will have to be changed anyway, since this is very
fundamentally not thread-safe and inefficient.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
The version_printed variable is used to print version information
only once. It was a global variable, but never used outside of
its function. Therefore, move it into the function and make it
static. Since this is a plain old datatype (POD), it makes no
no difference whatsoever whether the static variable is in block
scope or not. Indeed, it is initialized in the data segment). Well,
we are in C mode and therefore everything has to be POD by definition.
I tested this on gcc and clang.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Implement tag- and people-filtering in the mobile version of
DiveFilter. As opposed to the desktop version, this has no
different modes: it always searches "startswith" and "all of".
I.e. all of the search strings must match and a tag / person
is considered as matching if it starts with the search term.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
This function checked a dive for a search string. Its functionality
was replaced by a fulltext index.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
In analogy to the desktop version, use the fulltext index in
DiveFilter. This code is not yet executed.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
This function did not access any class members and was not used
outside the tranlation unit. Let's make it local (i.e. static)
to the translation unit.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
There are now three filter modes:
1) Dive site
2) Fulltext
3) Normal
When doing a fulltext search, get the dives that match the
fulltext filter and then apply the other filters on that list.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
When loading dive data, populate the fulltext index. When clearing
dive data, free the fulltext index. When deleting a dive, remove it
from the fulltext index.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Add code that indexes all words of a dive and provides searching
for words.
A query is represented by the FullTextQuery class, which can be
initialized by assigning a string to it. It is basically a list
of words.
The result of a search is stored in the FullTextResult class,
which is a list of dives.
The actual indexing and searching is implemented in the FullText
class. However, this class is not exported because the interface
is partially accessible to C. Notably, the reloading of the
fulltext index is done from the C core.
Currently, the indexing and searching is totally unoptimized.
In a ~4000 dives test-log searches typically took single-digit
ms times. There is ample room for optimization (e.g. when
searching for multiple words, chose the words with few dives
first and when down to a few dives, check them individually).
The words of each dive are tokenized and uppercased and
cached with the dive. A pointer to these words is stashed
in the dive structure.
For now, compile only on desktop.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
The FilterData struct has the enum StringMode, which describes how
strings are searched (substring, startswith, exact). To make it
more generally accessible, remove it from the class. Since it is
an "enum class", the values don't pollute the global namespace anyway.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
The DiveFilter class defined the showDive() function to test
whether a dive should be filtered or not. This was used in
DiveTripModel to loop over all dives or all dives affected by
an editing action.
This restricts us in how we do filtering: We can't use indexes
that give us directly the result. To make the filtering more
flexible, move the actual loops that do the filtering to
the DiveFilter class.
The undo-commands likewise called directly the showDive()
function to check whether newly added dives are shown.
Use the new interface here as well.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Because of the multiple string confusion, we'd get the names wrong for
modechange events. If we then made other changes to the dive and saved
the end result back, they'd now be wrong in the git cloud storage too.
Fix it up manually by just noticing that there's a 'divemode' string on
the event line, which can only happen with modechange events.
Maybe we should report the fixup? This just silently fixes it (but only
for the git save format).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We had some fairly obscure rules for how strings were parsed, and it
actually caused bugs when the same line had multiple strings in it.
That normally doesn't happen, and the cases where it was _supposed_ to
happen had special cases for it (divecomputer ID lines, and tag lines).
But by mistake, we had introduced a case of that for the event line
handling in commit b9174332d ("Read and write divemode changes (xml and
git)"), and nobody realized that the divemode string addition meant that
"oops, now it's corrupting the event name". An event line could look
like this:
event 40:00 type=8 divemode="OC" name="modechange"
where we now had both that "OC" and "modechange" strings, and the code
to pick the name just picked the first string. So we'd end up
effectively mis-parsing the above line as
event 40:00 type=8 divemode="OC" name="OC"
which is obviously wrong.
The dive mode didn't really need to be a string in the first place
(there is nothing to quote, and no spaces in it), but hey, here we are.
We can't just magially fix the existing broken saves.
So make it more straightforward to handle strings in the git format line
parser. We still stash the different decoded strings together in one
special memory buffer, but now the parser helpers automatically untangle
it as they traverse the key value pairs.
This is still overly subtle code, and it doesn't fix the cases where
we've saved the wrong data back. That comes later.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If we change the gps location of a dive that didn't have a dive site associated
before (which is the normal case when a dive was just downloaded from a dive
computer), a new dive site is created with that GPS fix and added to the dive.
We need to mark that dive as changed in order for the changes to be saved to
storage.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Currently, we do substring search. Implement starts-with and
exact mode (for example when search for "Cave vs. Cavern" tags).
For each textual search criterion add a combo-box.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Corrected typo of the word celsius in three files:
core/import-csv.c
core/divefileter.h
mobile-widgets/qml/Settings.qml
These were spelled as celcius but corrected these to celsius.
The 'core files were just comments but the mobile-widgets file would be
'active' code.
Reported by: tormento <turment@gmail.com>
Signed-off-by: Jason Bramwell <jb2cool@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
These just make no sense. Since the value is copied, it
has no meaning to the caller whether the function can
change the value (and vice versa for return types).
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Remove two erroneous comments stating that a function-local
QSettings variable should not be static because it is initialized
too early. Scoped static variables are initialized when execution
first hits the statement.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
The idea is that in portrait mode we can force the display to be single column (which
makes sure that the profile in dive display mode is nice and big).
This commit only implements the preference variable that we need for that.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Instead of transporting the global first and last dive date
in the dive summary, calculate it in an external function.
Since we already have time and date functions in qthelper.cpp
implement those functions there. Provide a stub in QMLInterface
so that QML can access these standalone functions.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Some Wikipedia pages use special (non-ASCII) unicode symbols for
representing the " and ' separators. Before parsing, replace these
by the ASCII symbols to enable copy & paste from Wikipedia (and
other sources?).
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
As we do XSLT parsing for the CSV import, ampersand characters need to
be encoded with & for the parsing to succeed.
Fixes#2037
Signed-off-by: Miika Turkia <miika.turkia@gmail.com>
