Re: [Tuxpaint-devel] Is this warning harmless?

[Bill K. <nb...@so...>]

On Wed, Nov 03, 2021 at 09:27:36PM -0700, Bill Kendrick wrote:
> On Tue, Nov 02, 2021 at 09:44:39PM +0900, TOYAMA Shin-ichi wrote:
> > Hi,
> > 
> > I think the reason for this warning is that the size of wchar_t is
> > 4 bytes on linux but 2 bytes on windows.
> > 
> > I don't know what render_text_w() is doing, so I'm not sure if
> > this is the problem or not.
> > 
> > For example, the text circled in red in onscreen_keyboard (see 
> > attached) does not seem to be rendered on windows.
> > 
> > Does this seem relevant ?
> 
> Hrm, we have some clues about this in the code from John P. from
> way back in 2006:
> 
>   https://sourceforge.net/p/tuxpaint/tuxpaint/ci/304f49672/
> 
> Specifically, this part of the commit message:
> 
>   Modified the text gadget so that it correctly handles the 16-bit unicode
>   characters that SDL sends. The text buffer is held internally as an array
>   of wchar_t, and makes uses of various wide-character functions.
>   It is converted back into 16-bit unicode characters to satisfy SDL_ttf.
>   Tested on Windows and Linux.
> 
> And it's this function:
> 
>   // This conversion is required on platforms where Uint16 doesn't match wchar_t.
>   // On Windows, wchar_t is 16-bit, elsewhere it is 32-bit.
>   // Mismatch caused by the use of Uint16 for unicode characters by SDL, SDL_ttf.
>   // I guess wchar_t is really only suitable for internal use ...
>   static Uint16 *wcstou16(const wchar_t *str)
>   ...
> 
> It is used in exactly one place, within render_text_w():
> 
>   if (font->typ == FONT_TYPE_TTF)
>     {
>       ustr = wcstou16(str);
>       ret = TTF_RenderUNICODE_Blended(font->ttf_font, ustr, color);
>       free(ustr);
>     }
> 
> 
> However, we're talking about the bit of code that constructs
> UTF-8 Unicode within Tux Paint, to deliver it to SDL_Pango.
> 
> I think it might be worth replacing `wchar` with something
> like Uint32 for these purposes...

Ugh, so another problem I'm seeing is we actually store strings
in `wchar` form as the Label tool's text inside the PNG images.

I assume that, because of this, Tux Paint pictures drawn on
Linux might not work under Tux Paint on Windows, and vice-versa,
due to the different sizes (16- vs 32-bit).



Back to the issue at hand, it actually seems like anything about
0xFFFF isn't really useful to us.

  https://unicodebook.readthedocs.io/unicode.html#unicode-categories

says:

  There are 3 ranges reserved for private use (Co subcategory):
  U+E000 - U+F8FF (6,400 code points), U+F0000 - U+FFFFD (65,534) and
  U+100000 - U+10FFFD (65,534). Surrogates (Cs subcategory) use the
  range U+D800 - U+DFFF (2,048 code points).

I'm thinking it's probably safe to just ignore str[] values above
0xFFFF (which are impossible on environments, like Windows, with
16-bit `wchar`).

I'll see what happens on Linux.  If it seems okay, I'll commit, and
you can test things on Windows.


Regarding those characters that did not appear in the OSK for you,
I made the above discover by simply replacing the code that converts
any >= 0x00010000 values in str[] to UTF-8, for sending to SDL_Pango,
and just doing this:

  utfstr[j++] = 'X';

...But I could still see the symbols (and not an "X"). :)

Perhaps that's a font issue?  Can you play around with it?


-bill!


> 
> I'm digging around the code right now.
> 
> -bill!
> 
> 
> > 
> > 
> > TOYAMA Shin-ichi wrote in <617e94b2.5007%sh...@wm...>
> > >Hi,
> > >
> > >Recently, I've been dealing with a lot of bugs in the Windows
> > >version, and I've noticed that many clue of the bugs had been 
> > >appeared in the compile-time warnings.
> > >
> > >So,I checked the warnings again, and the following warning,
> > >which does not appear when compiling on Linux, caught my 
> > >attention.
> > >
> > >------------------------------------------------------------
> > >src/tuxpaint.c: In function 'render_text_w':
> > >src/tuxpaint.c:1685:27: warning: comparison is always true due to limited range 
> > >of data type [-Wtype-limits]
> > > 1685 |           else if (str[i] <= 0x0000FFFF)
> > >------------------------------------------------------------
> > >
> > >If this warning is correct, it means that the next conditional
> > >branch (0x00100000 - 001FFFFF) will never be reached.
> > >
> > >How can I confirm if it were a problem or not?
> > >
> > >Thanks!
> > 
> > -- 
> > TOYAMA Shin-ichi mailto:sh...@wm...
> 
> 
> 
> > _______________________________________________
> > Tuxpaint-devel mailing list
> > Tux...@li...
> > https://lists.sourceforge.net/lists/listinfo/tuxpaint-devel
> 
> 
> -- 
> -bill!
> Sent from my computer
> 
> 
> _______________________________________________
> Tuxpaint-devel mailing list
> Tux...@li...
> https://lists.sourceforge.net/lists/listinfo/tuxpaint-devel

-- 
-bill!
Sent from my computer