[Ohrrpgce] Android SdlVideoResize=y

[James Paige]

On Sun, Jun 09, 2013 at 02:53:40PM +1200, Ralph Versteegen wrote:
> On 9 June 2013 05:55, James Paige <Bob at hamsterrepublic.com> wrote:
> > On Sun, Jun 09, 2013 at 05:42:16AM +1200, Ralph Versteegen wrote:
> >> On 9 June 2013 04:00, James Paige <Bob at hamsterrepublic.com> wrote:
> >> > On Sat, Jun 08, 2013 at 09:01:20PM +1200, Ralph Versteegen wrote:
> >> >> On 8 June 2013 02:26, James Paige <Bob at hamsterrepublic.com> wrote:
> >> >> > On Fri, Jun 07, 2013 at 09:12:35PM +1200, Ralph Versteegen wrote:
> >> >> >> On 7 June 2013 03:05, James Paige <Bob at hamsterrepublic.com> wrote:
> >> >> >> > On Thu, Jun 06, 2013 at 07:52:15AM -0700, James Paige wrote:
> >> >> >> >> On Thu, Jun 06, 2013 at 01:35:03PM +1200, Ralph Versteegen wrote:
> >> >> >> >> > On 6 June 2013 04:53, James Paige <Bob at hamsterrepublic.com> wrote:
> >> >> >> >> > > I was experimenting with the Android build a little. It runs really nice
> >> >> >> >> > > on the OUYA. Don't Eat Soap goes the full 18.2 fps, and bell of CHaos
> >> >> >> >> > > runs at a pretty solid 18.1 fps (on the first screen with no enemies, I
> >> >> >> >> > > couldn't jump)
> >> >> >> >> > >
> >> >> >> >> > > I change the button mappings. I figured there was no point to mapping
> >> >> >> >> > > any of the buttons to arrow keys because there is already an arrow key
> >> >> >> >> > > thingy on the left, so just for testing Bell of Chaos I mapped:
> >> >> >> >> > >
> >> >> >> >> > > AppTouchscreenKeyboardKeysAmount=4
> >> >> >> >> > > AppTouchscreenKeyboardKeysAmountAutoFire=0
> >> >> >> >> > > RedefinedKeysScreenKb="LCTRL LALT RETURN ESCAPE"
> >> >> >> >> > > RedefinedKeysScreenKbNames="CTRL ALT Enter ESC"
> >> >> >> >> > >
> >> >> >> >> > > and tested it on a Nexus 7 at work. It played very nicely. Small
> >> >> >> >> > > slowdown when a lot of stuff is happening, but still totally playable.
> >> >> >> >> >
> >> >> >> >> > Cool!
> >> >> >> >> >
> >> >> >> >> > > One thing I have noticed on all three of the devices I have tested is
> >> >> >> >> > > that the screen only takes up the top left corner, and that there are
> >> >> >> >> > > black spaces on the right and bottom, like this:
> >> >> >> >> > > http://i.imgur.com/O7F3NxB.jpg
> >> >> >> >> > >
> >> >> >> >> > > I tried setting:
> >> >> >> >> > >
> >> >> >> >> > > SdlVideoResize=y
> >> >> >> >> > >
> >> >> >> >> > > But it seemed to make no difference. (and I did remember to increment
> >> >> >> >> > > AppVersionCode)
> >> >> >> >> > >
> >> >> >> >> > > ---
> >> >> >> >> > > James
> >> >> >> >> >
> >> >> >> >> > If you want to try stretching the screen, modify gfx_sdl.bas.
> >> >> >> >> > Currently it asks SDL_SetVideoMode for a buffer the size of the
> >> >> >> >> > screen, and then uses the highest zoom amount possible. I did that
> >> >> >> >> > because on my emulator when I only asked for a 320x200 screen it was
> >> >> >> >> > glitchy with the image cut up into pieces.
> >> >> >> >>
> >> >> >> >> Oh, I understand!
> >> >> >> >>
> >> >> >> >> I/OHRRPGCE(20248): gfx_sdl: screen size is 1280*736
> >> >> >> >> I/OHRRPGCE(20248): gfx_sdl: selected zoom = 3
> >> >> >> >>
> >> >> >> >> So SDL_SetVideoMode(0, 0, bitdepth, flags) returns the actual
> >> >> >> >> resolution, 1280*736 and then allocates a screen of that full size. Then
> >> >> >> >> select_zoom_automatically finds the biggest zoom that fits within that
> >> >> >> >> size.
> >> >> >> >>
> >> >> >> >> ... *testing*
> >> >> >> >>
> >> >> >> >> I forced the screen size to 320x200 and zoom to 1, and then recompiled
> >> >> >> >> with SdlVideoResize=y and it works beautifully on the Nexus 7
> >> >> >> >>
> >> >> >> >> http://i.imgur.com/Rsghv3v.jpg
> >> >> >> >>
> >> >> >> >> Now if we can just figure out a method that works on both old and new
> >> >> >> >> devices, or a graceful way to pick between the two screen sizing
> >> >> >> >> methods.
> >> >> >> >
> >> >> >> > Just a thought on this, what if we leave SdlVideoResize=y in the
> >> >> >> > AndroidAppSettings.cfg and then make gfx_sdl respect the -f command line
> >> >> >> > option on android to do a 320x200 zoom=1 screen? but when -f is not
> >> >> >> > specified, it will continue to use the current behavior (although we
> >> >> >> > probably still do want to center it rather than top-lefting it)
> >> >> >> >
> >> >> >> > ---
> >> >> >> > James
> >> >> >>
> >> >> >> I just tried out turning on SdlVideoResize myself, and it works fine
> >> >> >> on both my phone and all emulator instances. I don't know what caused
> >> >> >> the problem which I experienced before.
> >> >> >
> >> >> > Ah, wonderful! :)
> >> >> >
> >> >> >> I'm surprised that I can barely notice the blurriness due to the
> >> >> >> stretching and linear interpolation, even on text. (It's really
> >> >> >> noticeable when you take a screenshot and scale it up to 200% though).
> >> >> >> However I find the change in aspect ratio when 320x200 isstretched to
> >> >> >> 320x240 surprisingly noticeable; I can immediately tell that tiles
> >> >> >> aren't square and NPCs are skinnier. Of course the DOS versions of the
> >> >> >> OHR used to have exactly the same amount of stretching when played on
> >> >> >> the common monitors of the day ;).
> >> >> >
> >> >> > Yeah, I remember back in the DOS days I expected tiles to be non-square,
> >> >> > and just didn't mind because I knew the pixel count. After the Windows
> >> >> > port, for a while, running in Windowed mode felt strange to me because
> >> >> > it seemed vertically squished :)
> >> >> >
> >> >> >> So it would be nice to have an option to preserve aspect ratio (though
> >> >> >> I would still have it off by default I think). SDL's "perserve 4:3
> >> >> >> aspect ratio" setting does just that: it causes stretching to a 4:3
> >> >> >> ratio, rather than preserving the aspect ratio of the resolution the
> >> >> >> app actually asked for. (You can turn the option on and off in the
> >> >> >> config menu without having to recompile). We could either patch the
> >> >> >> SDL fork, or put our own black bars on the sides (calculated with
> >> >> >> knowledge of the real resolution) to preserve the ratio.
> >> >> >
> >> >> > Yeah, that sounds good as long as we can switch it at runtime as opposed
> >> >> > to compile time. What is this config menu you mention? Does the
> >> >> > sdl-android port have its own built-in config menu? And if so, how do I
> >> >> > access it?
> >> >> >
> >> >> > Oh, this also reminds me, what about summoning the on-screen keyboard. I
> >> >> > see there are some commands related to this in
> >> >> > project/jni/sdl-1.2/include/SDL_screenkeyboard.h I think we can use them
> >> >> > in gfx_sdl as long as we wrap them in #ifdef __FB_ANDROID___
> >> >> >
> >> >> > I was thinking we should have a request_virtual_keyboard()
> >> >> > hide_virtual_keyboard() pair of commands which would do nothing on most
> >> >> > platforms, but on android would open/close the viertual keyboard.
> >> >> >
> >> >> > Then in game, we could call those functions when a hero is re-named, and
> >> >> > when the "input string" plotscripting command is used.
> >> >> >
> >> >> > ---
> >> >> > James
> >> >>
> >> >> Yes, that's a good idea.
> >> >>
> >> >> This would be the first instance of patching a game's scripts. We'll
> >> >> have to replace any script named "input string" with the
> >> >> android-specific version. Doing so based on checksum isn't really
> >> >> possible because the script calls other scripts, which have variable
> >> >> script IDs. (scanscripts.py shows 38 different checksums).
> >> >
> >> > I think it would be best to just hook up the android keyboard trigger to
> >> > the "enable input text" command. I know that command is only about a
> >> > year old, so it would mean that people who used "input text" in their
> >> > scripts would have to re-compile them before publishing on android.
> >>
> >> But that's not what "enable input text" does at all. I don't know if
> >> anyone has actually used this command, but you're meant to call it at
> >> the beginning of the game, not at the point that you call "input
> >> text"... in fact, doing that won't work. So I think it's something
> >> that should be a new command if added at all, but that changing
> >> inputstring is better.
> >
> > Oh, I guess I misundersttood that command. It is definitely used by
> > the latest iplementation of "input text"
> 
> (You mean "input string". Also, in any previous emails in which I said
> "input text" I meant "get input text"). Oh, I forgot that I had
> actually used it there.
> 
> >> > But that seems better than a really complicated solution.
> >>
> >> If you mean "patching" scripts, that's not complicated at all. Iterate
> >> through the scripts when loading the game, find one named
> >> "inputstring", provide an alternative compiled .hsz file.
> >
> > Ah, well if that is easy, then I can't possibly complain ;)
> >
> > Seems like that would only work for patching scripts that only call
> > built in functions, and do not call any other scripts in plotscr.hsd
> >
> > Aren't the id numbers for compiled scripts allocated dynamically at
> > compile-time?
> >
> > ---
> > James
> 
> Are you talking about replacing script based on checksum? As I said,
> because the ID numbers are variable the only realistic solution is to
> replace based on script name.

Nah, I understood what you meant about replacing by name.

I was thinking about references to a compiled script from a compiled 
script. I don't think that is actually an issue with "input string", I 
was just wondering about how scripts that call each other store the 
reference to the script that they are calling.

I know in the older implementations of hspeak, scripts called each other 
using autonumbered id numbers, and those could change if you recompiled. 
I don't know if that is even still the case anymore.

---
James