Passing segment instances fails because this creates a race condition.
When a long conversion is taking place, the SignalBase::samples_added
signal is called often but since it's in a separate thread, the calls
are queued and aren't executed immediately. Now if the conversion is
restarted - for example as a result of a changed conversion threshold -
then the segments holding the converted data are destroyed, rendering
the pointers submitted as parameters to samples_added invalid.
Once the signal queue is processed, those invalid pointers will be
accessed and PV segfaults.
Since the signal queue can neither be emptied nor flushed, this
leaves only two sensible choices:
1) Signal samples_added less often, thereby reducing the chance of
signals being queued
2) Supply the segment ID instead of the segment instance as that's
essentially the only thing we currently care about - in fact, the
only user of samples_added (ViewBase::on_samples_added) uses the
instance to query only this
As #1 is only a band-aid and not a waterproof solution, I chose
to go with #2.
In file included from [...]/pv/globalsettings.cpp:20:0:
[...]/pv/globalsettings.hpp:31:12: error: 'std::function' has not been declared
using std::function;
^
test/CMakeFiles/pulseview-test.dir/build.make:88: recipe for target 'test/CMakeFiles/pulseview-test.dir/__/pv/globalsettings.cpp.obj' failed
make[2]: *** [test/CMakeFiles/pulseview-test.dir/__/pv/globalsettings.cpp.obj] Error 1
make[2]: *** Waiting for unfinished jobs....
[ 5%] Building CXX object CMakeFiles/pulseview.dir/pv/globalsettings.cpp.obj
In file included from [...]/pv/globalsettings.cpp:20:0:
[...]/pv/globalsettings.hpp:31:12: error: 'std::function' has not been declared
using std::function;
^
Add #include <functional> everywhere where std::function is used. Only
the occurence in globalsettings.hpp was causing an actual build failure
on MinGW, though.
The previous implementation used a "*" filter when a file of any other
format than srzip got imported. This happened to be a duplicate of
"All files", and ignored the list of filename extensions provided by
the input formats.
This change does respect the input format's file extensions, and copes
with the lack of such a list (raw binary), as well as lists that have
one (most formats), or multiple extensions (raw analog).
This fixes bug #1039.
Another byproduct of the change is that extensions and their decoration
(separators, parentheses) move outside of tr() calls. These technical
details shall not concern translators, and translations for human
languages shall not break the filter mechanism.
This implementation might be "too complex, computationally expensive".
But it works, and fixes an issue, and the code path executes seldom and
waits for user interaction anyway. Cost reduction can get applied later.
Up to now, registered callbacks could not be unregistered
because std::function does not permit comparing for equality.
Using an interface class removes the need for std::function,
making the mechanism a little less elegant but at least fully
functional.
As reported in bug #1118 not all input sources provide a samplerate, and
decoder instances may not cope with a rate spec of 0. Only pass non-zero
sample rates to the decoder stack. This improves robustness in addition
to the specific fix for #1118 in the decoders' implementations.
Add a -D command line option which skips auto-detection of devices upon
startup. This can speedup program startup for setups with known devices,
and allows to skip the scan for troubled drivers which might break the
startup phase.
This -D option can be combined with -d specs, either presence or absence
of -d is acceptable when -D is specified. Users still can interactively
scan for and use devices after program startup.
This resolves bug #1116.
Use case is as follows:
- Capture 20+ segments with ~500kS each
- Afterwards, enable conversion for a channel
Without this change, the converted logic will be repainted
20++ times because we are only told that new samples were
added but not which segment.
With this change, the logic trace is only painted when we
see that samples were added to the segment we're showing.
Instead of terminating, we wait instead.
We do this because SignalBase::on_samples_added() somehow doesn't
reliably see that there's no conversion thread active anymore.
conversion_thread_.joinable() returns true when the thread was
already terminated for whatever reason, resulting in on_samples_added()
trying to notify a non-existant thread.
Instead of handling the metadata separately from the mux/decode segments,
it's much neater to handle both together. Doing this also allows us to
remove the need for query_input_metadata() since we're taking the metadata
from the muxed logic segments.