[Freeciv-Dev] (PR#2922) Re: (PR#2924) OK. the reorg patch has a bug. Help from a client expert?

["esr@xxxxxxxxxxx via RT" <rt@xxxxxxxxxxxxxx>]

Jason Short via RT <rt@xxxxxxxxxxxxxx>:
> The file graphics/isotrident/terrain1/tx.village.png is incorrect. 
> Looking at isotrident/terrain1 I don't see how this output could have 
> been reached.  But the end result is that tx.village for isotrident 
> seems to have copied that of trident.

I looked in the isotrident and trident tile arrays.  The village icons
in these look identical to me. Is it possible that you might have been
misled by the visual resemblence, or is it me that's being blind?

In any case, if I *have* made this error, it should have resuulted in getting
the wrong icon in the right place, rather than a correct-looking icon in the
wrong place :-).

No.  There's something hosed in the calculation of offsets.  

> There is a one-pixel boundary between grid elements if is_pixel_border 
> is set.  For instance in tilespec.c:
> 
>       x1 = x_top_left + column * dx + (is_pixel_border ? column : 0);
>       y1 = y_top_left + row * dy + (is_pixel_border ? row : 0);
> 
> Perhaps you are not taking this into account?

Um.  I don't think I messed this up.  But maybe I'm missing something.

Remember that the sprite loader is not looking into a tile array any
more; the 0, 0 corner of the image is the 0, 0 corner of the sprite.
In the new system, is_pixel_border is always zero.

You know what it looks like to me?  It looks like the drawing routines
think the size of every sprite is the base size of the tileset.  So
the full-width/full-depth tiles in the terrain set look OK, but the
small ones get tucked into the upper-left-hand corner of a rectangle
of the tileset's base size centered on the correct location.

> A bigger problem is that the new graphics are 8x larger than the old 
> ones.  On my computer the minimum block size is 4k, which is what causes 
> this (most files are 1k in size).  But it doesn't seem good...

True.  But Daniel Speyer pointed out, on the other hand, that the new
organization can be used to decrease startup latency by implementing
load-on-demand.  Trading some disk storage for snappier response seems
like a good thing.
-- 
                <a href="http://www.tuxedo.org/~esr/";>Eric S. Raymond</a>