This addressee a bug where ardour 6 was able to write negative
duration `length="-1"` `length-beats="-3.3650500597559585e-05"`
Ideally timecnt_t::string_to should check for invalid,
negative, duration. But this also catches a more generic case.
```
exception at str.substr (1)
#3 Temporal::timepos_t::string_to (this=0x7fffffff7bb0, str="") at libs/temporal/timeline.cc:904
#4 Temporal::timecnt_t::string_to (this=0x7fffffff7ba0, str="-2") at libs/temporal/timeline.cc:294
#5 PBD::string_to<Temporal::timecnt_t> (str="-2") at libs/ardour/ardour/types_convert.h:131
```
The default clock-limit is 99:59:59:00, just under 360000 seconds
(see ARDOUR_UI::parameter_changed, clock-display-limit).
AudioClock calculates this limit pos as
`timepos_t (limit_sec * _session->sample_rate())`
This caused an overflow leading to a negative value:
```
timepos_t (359999 * 96000)
samples_to_superclock (359999 * 96000, 96000)
int_div_round (359999 * 96000 * 282240000, 96000)
```
Ideally this will be optimized, here the sample-rate cancels out,
so we could use a c'tor usin seconds.
In other cases we could cache the pre-calculated sc_per_sample:
`superclock_ticks_per_second() / superclock_t (sr)` which is an
integer for all commonly used sample-rates.
Before, the X-axis labels would be placed in nice even positions, but
with whatever odd time corresponded to that. That made it harder than
necessary to read the graph and approximate when things happened.
Instead, round the interval down to nearest power of ten ... and if
suitable, scale that up with a factor of 2 or 5. That will (with the
necessary handling of how seconds/minutes/hours relate) make sure that
the time labels are nice with a minimal amount of non-zero digits. That
makes it easy to do math and interpolate when reading the graph.
The number of labels will change between something like 4 and 10 -
before it was always something like 7. That is fine, as long as it helps
the readability.
The total length is no longer the right-most label, but the length can
be found in the top-right corner.
This moves the _amp from send to delivery (which already
applies gain for the master-bus normalization). This generalizes
the use of a gain stage for use in port-inserts.
Previously active Routes were retained until the end of
Session d'tor and not dropped during Session::destroy.
While most resources were explicitly cleaned up via DropReferences,
Processor UIs are kept around until the actual destructor runs.
Likewise some controllable are kept around while the GraphNode (Route)
owning it is not released.
* GTK fills in the snapshots in the background (using background threads
to test files).
* Actual colors may only available after the widget is realized (not
directly after ensure_style)
It can happen that with a scaling factor of 1.0, rubberband
produces slightly fewer samples than the original.
Region::set_length (region->length * 1.0) is idempotent and
does not shorten it as appropriate to the longest source
via Region::verify_length(), which leads to various issues.
* Allow to bundle presets
* Do not use /usr/[local] on Windows
(This may need further work, for Windows, since default user
presets are in `file://$HOME/.ladspa/rdf/`)
