I'm not sure if this test is going to be effective as I don't have hardware to
test on at the moment. As noted in the documentation, Windows XP should be the
only OS where QPC uses a timer source that is non-monotonic(multi-core with
non-syncronized TSC).
timeEndPeriod must be called with the same timer resolution value used in
timeBeginPeriod. When the process exits the timer resolution is restored anyway
so this is not very important.
Add functions for get/set the Multimedia timer resolution, although we are
really only interested in the minimum, this will facilitate testing
Put timer utility functions inside nested namespaces as they are platform
specific
Evoral::Beats::operator>() rounds to (1.0 / PPQN), hardcoded 1/1920.0.
If the time difference between two events is smaller than 1/PPQN,
Beats::operator>() and Beats::operator<() produce ambiguous results.
The same pair of values is both "less than" and "greater than" depending
which operator is used.
While it's fine for some cases to ignore the order of nearly concurent
events, the std::priority_queue must be strictly ordered.
We should still prevent a user from naming a route "Monitor", but
since "Monitor" is a translatable string, this is an i18n safeguard. eg.
- create a session with a mon-section and a track Монитор.
- send the session to a friend in Russia.
This is a first approximation. Doing this 100% efficiently is quite hard because it would require comparing the range
currently in the disk buffer and the range impacted by the layer change. I suspect this will work fine for the
foreseeable future
This is akin to VST2's audioMasterGetBlockSize.
It returns the current nominal block size (think jack-buffersize).
It's not the only block size that may be used when calling run(), it's
just the normal one. The actual block sizes used may be larger or smaller
and may vary between successive calls of run().
This change became neccesary after 53e969e9. Some plugins expected
maxBlockLength to be the /current/ buffer-size and not all-time maxiumum.
Those plugins can now use nominalBlockLength.
- copy old peak-files to new (do not require re-calc)
- keep old peak-files (for now, backwards compat)
- fix cleanup-sources to remove *new* peak-file
- include channel-number in hash (like it was done before)
see also 624f76b
TODO: add Session > Cleanup > remove/re-create peaks
Ardour-2.0 put peak-files in a "broken" location for several months.
Since then Ardour renamed those files. No more. When loading ancient
sessions peak-files are now re-created (in the background).
The original broken code was intended to stop using ifstream
and ofstream to open files, but it used a very old MSVCism to
open streams from FILE*.
In the case for reading, this patch replaces that code by using
standard c file manipulation to read the data into a stringstream,
and then it uses the existing stream-based code to read from there.
Similarly the ofstream in the code for writing the recent files list
is replaced by a stringstream, and then a c-string is extracted
from it and written using standard c file manipulation.
Also tweaked a couple of typos from rgareus :)
Use gdk_pango_layout() for all text-size calculations.
In Canvas::Text always use the same context (local image surface)
for layouting (prior to this, local and window-context mix resulted in
different sizes).
On linux SndFileSource::get_soundfile_info was being called with an empty path
value when clicking on a folder in the left side part of the file browser which
resulted in an unnecessary error message.
On Windows every time a directory was selected in the Import dialog,
SndFileSource::get_soundfile_info would fail to open the directory in read only
mode and produce an error, but as we don't want to query/open soundfile info
for directories anyway just test and return if the path is a directory.
