Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
Working on Vulkan clarified what has always been slightly smelly about the
design and organization here: not everything that is API specific is really in
a "backend" (a PuglBackend). The concrete example is puglGetProcAddress(),
which only makes sense for GL and is actually implemented in the "backend"
files. Arguably puglGetContext() is also such a thing.
So, rename the headers so they can be the place where API-specific things go in
general, which happens to include a backend most of the time. The stub is a
bit of an exception to this, but whatever. The includes look tidier this way.
In place of the old headers are compatibility stubs that just emit a warning
and include the new version, which will be maintained for a while.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
It turns out that there are some systems where having this on makes things not
work, because compatible profiles are only supported up to earlier versions,
despite a much higher version being available.
|
|
|
|
|
|
Currently unused, but this is to leave open the possibility of event
propagation or better errror handling.
|
|
|
|
This finishes the removal of the init/set split. While these ones are
superficial, the general idea here is to provide general functions that work
before or after window creation where possible. This prevents the situation
where ever more dynamic counterparts to existing init functions get added over
time.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This allows waiting for events for any view in the world. It also improves on
puglWaitForEvent() by the addition of a time parameter that allows indefinite
blocking, non-blocking polling, and blocking polling with a timeout.
|
|
|
|
The old API was broken for programs that manage multiple views, since it was
impossible to wait for events on any view. There are also several functions in
the API which are not actually associated with views at all, so those can now
be moved to the more appropriate PuglWorld to make this more clear.
The old puglInit() and puglDestroy() functions are preserved for compatibility,
but marked as deprecated.
|
|
|
|
Not very becoming of a library that exists specifically to avoid static data.
|
|
|
|
|
|
|
|
|
|
|
|
It's possible that this does bad things if a UTF8 string is used on a system
that does not support it (because XStoreName is still called), but I have no
idea how likely this scenario is.
Leaving it this way because it means the ASCII case will still work everywhere,
and it's easy enough to avoid if this is a problem.
|
|
Mainly to be able to test backends, this is still a silly program.
|
|
|
|
This removes PuglContextType and allows the user to pass a backend directly.
Normally this would come from one of the two accessors declared in the headers,
but it would be possible to add backends without changing any of the existing
code at all.
Unfortunately, it is not possible to preserve backwards compatibility and
achieve the ultimate goal of linking only to the required dependencies, so
puglInitContextType() has just been removed.
|
|
|
|
These need to be symmetric because sometimes different things need to happen in
either situation when entering the context as well.
|
|
|
|
|
|
There are two possible approaches to take here: try to expand dimensions that
are not being explicitly resized (for example expand the bottom when dragging
right), or just stop single-dimension resizes if they would go out of range.
I chose the latter here for two reasons: it's hard to always do something
smooth and unsurprising with the first approach (and it would require more
code), and it can be nice from the user's perspective to easily be able to
resize the window to exactly one of its aspect ratio limits. For example, it
is very easy to drag pugl_test to 1:1 or 16:9.
In other words, simplicity and user power wins.
|
|
Weird convention, that.
|
|
Only one field is necessary to store any kind of key, including special keys,
since PuglKey occupies a reserved Unicode region. This is generally much
simpler to deal with since there is only one value to dispatch on.
Text events are separated from key events (like Windows but unlike MacOS or
X11) because it is not possible to derive text events from key press events
when they occur on Windows. Since merging the two has been the source of some
confusion, this approach has some advantages anyway, even though it introduces
the need to handle another event type.
In the process, text input has been almost completely rewritten. I have tested
this with a compose key on X11 and dead keys on Windows and MacOS and
everything seems to work correctly, though there may (as always) still be
issues with more exotic input methods.
|
|
This matches Gtk and is generally less annoying to work with.
|
|
|
|
|