Age | Commit message (Collapse) | Author |
|
No idea why they were bools in the first place.
|
|
|
|
|
|
It's just as good: at 30 FPS, this allows for
2 ^ 32 / 30 / 60 = 2386093 mins, which is, I guess, a reasonable limit.
|
|
|
|
|
|
|
|
|
|
Conflicts:
engines/wintermute/video/video_subtitler.cpp
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Conflicts:
engines/wintermute/video/video_subtitler.cpp
|
|
|
|
|
|
|
|
This should fix bug #6720.
|
|
|
|
|
|
|
|
Signed/unsigned comparison.
|
|
|
|
This reverts commit 5e099e90bcdfb87c289b919ddac60e60fbf6bdd7.
There's no need for this addition, since the intro was skippable before that too, by hitting the ESC key.
See System::touch().
|
|
This was the bug with "Engine assertion at various dialogues if no items
in inventory".
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Because of the way the speech is compressed with duplicate samples
being stored with a negative size and a single value, when reading the
data with the wrong endianess we can end up with a lot of duplicate
samples which biased the result with the way the old heuristic was
coded. Hopefully this change to skip duplicate samples will make it
more robust.
|
|
Because the data is compressed (a repeated sample is coded as a
negative length followed by the value), when the length is read with
the wrong endianess we get completely wrong data. So to get the BE
data we cannot just read them assuming LE and byteswap afterward.
|
|
|
|
|
|
|