- 02 Jun, 2021 2 commits
-
-
refs #25781
-
-
- 01 Jun, 2021 2 commits
-
-
ends up reading wrong sample size/time refs #25784
-
-
- 31 May, 2021 1 commit
-
-
- 29 May, 2021 16 commits
-
-
Lyndon Brown authored -
Lyndon Brown authored
-
Lyndon Brown authored
-
Lyndon Brown authored
the null entry should not be there. caught this with some work to be submitted later.
-
Lyndon Brown authored
these lists takes signed ints. this was caught by some later work that caused a warning here.
-
Lyndon Brown authored
-
Lyndon Brown authored
unused since 8f45b1e5
-
Lyndon Brown authored
such strings are supposed to be translated JIT for display by interfaces. this mistake causes a secondary translation to occur at an earlier and incorrect time - in execution of the plugin descriptor.
-
Lyndon Brown authored
such strings are supposed to be translated JIT for display by interfaces. this mistake causes a secondary translation to occur at an earlier and incorrect time - in execution of the plugin descriptor.
-
Lyndon Brown authored
'cat hints' should only be used by the core (to define a set of headings to break its large set of options up with, instead of the option tree items). incorrectly using a cat hint here instead of a section means that this "hacks" sub-heading gets printed like a module heading, which is both imperfect from a neatness point of view, but also and more importantly potentially confusing.
-
Lyndon Brown authored
usage hint is the wrong thing to use, help text is the right thing; this mistake is the cause of confusing help output due to the fact that usage hints are printed exactly the same as a plugin/module heading.
-
-
-
-
note that the line adding a category entry is because it's moved from above.
-
-
- 27 May, 2021 6 commits
-
-
VolumeSet (called from any threads) was accessing the au_unit variable, that is written and setup from the aout stream thread. This also fixes possible crashes when changing the volume after a failing Start().
-
-
Fix data race with au_unit. Indeed, the aout can be muted from any threads. SPDIF playback can now be muted.
-
Fix data race with au_init, leading to crashes, since the audiounit was stopped from a different thread than the playback one. Indeed, mute callback can be called from any threads.
-
It can be used as a hint to whether the buffer needs to be processed or not.
-
This will write 0s from the render callback without touching at the playback buffer state.
-
- 25 May, 2021 2 commits
-
-
FT_Glyph_Stroke is more tolerant of font files that do not follow contour orientation conventions, like Helvetica Neue. It also creates unfilled outline glyphs that can be more efficiently blended. Fixes #21043
-
Steve Lhomme authored
It is imprecise and on seek it may wait a lot of time as, while it's waiting, the decoder threads will flood new frames, delaying that much longer until the ID3D11DeviceContext is finished doing things. This workaround did more harm than good. Now that we have ID3D11Fence support (any recent Win10) the query based system should not be used. For older Windows using D3D11 we'll have less precise estimation of the time it took to actually render, meaning the frame drop will be less effective.
-
- 24 May, 2021 2 commits
-
-
-
closes #25588
-
- 23 May, 2021 9 commits
-
-
-
-
-
-
the parent is always the table.
-
we need 4 columns, not five. one element on the line with four was being pointlessly stretched across two.
-
now the above non-global check has been reduced to a single name check, we can strip the reverse of that here with an else. the one remaining condition is just to avoid work for empty mappings.
-
unlikely and better to add it rather than skip. being there with a blank label would better help diagnose what went wrong in testing, or if it slipped past testing, someone is more likely to notice and report the problem leading to a fix. note that we must conditionally use `qfut()` because the underlying translation function does not like nulls, and my fix for that was rejected.
-
we now rely upon checking option name rather than shorttext. it is somewhat less likely that a mistake with duplicate option names will have been made compared to duplicate shorttext, and duplicate option names presents an even bigger and broader problem that we currently do nothing about. this is not very useful towards helping with that, we'd want a full check of the entire table's option names if we cared to properly check things here.
-
