This bug report was migrated from our old Bugzilla tracker.

These attachments are available in the static archive:

• bug.c + PT Serif_Italic.ttf + Makefile (PT_Serif_Italic_Bug.zip, application/x-zip-compressed, 2019-02-28 15:40:50 +0000, 162762 bytes)

Reported in version: 2.0.15
Reported for operating system, platform: Linux, x86_64

Comments on the original bug report:

On 2019-02-28 15:40:50 +0000, ace wrote:

Created attachment 3660
bug.c + PT Serif_Italic.ttf + Makefile

Initially this bug was found under windows in my C# project used shared libs via wrapper SDL2-CS
SDL 2.0.8
SDL_ttf 2.0.15 / 2.0.14
After bug localization i could confirm it under linux in pure c(files in attachment)
Bug description:
If you use TTF_OpenFontIndexRW or TTF_OpenFontRW for creating TTF_Font with same SDL_RWops (for creating SDL_RWops I used SDL_RWFromFile, SDL_RWFromMem, SDL_RWFromConstMem) and you render text using TTF_RenderUTF8_Blended:

first TTF_Font font_2 with russian symbol "й"
second TTF_Font font_6 with any text: text will be blank(i set TTF_STYLE_UNDERLINE to second one to indicate that it was being rendered)
You got this:
https://drive.google.com/file/d/1Fo9JFQsPJuMTW6_FDJ1ZcZ-3K8IdgUCD/view

If you don’t use russian symbol "й" in first render, you will get this:
https://drive.google.com/file/d/1qgRupAG5SFobG3NXAM46j6ilvHx1STPt/view

Finaly, if you use TTF_OpenFontIndex / TTF_OpenFont instead of TTF_OpenFontIndexRW / TTF_OpenFontRW, rendering works as it should:
https://drive.google.com/file/d/1HP0RxDoSknUHHEHIfK_HN96SYesBYlOM/view

Hope you can help me, because using TTF_OpenFontIndex instead of TTF_OpenFontIndexRW is very expensive.

On 2019-02-28 16:31:06 +0000, Sylvain wrote:

You need to rewind the RWops between the two calls of TTF_OpenFontIndexRW:

SDL_RWseek(RWops, 0, RW_SEEK_SET);

But I'm not sure why there was no error ...

On 2019-02-28 20:45:38 +0000, ace wrote:

Thanks for your comment!
Yes, it was solve problem in c, but SDL_RWseek() is macro, and unavailable in bindings (https://www.libsdl.org/languages.php)

I think good solution will be:
Calls SDL_RWseek() in TTF_OpenFontIndexRW/TTF_OpenFontRW
AND/OR
Replace this macro with function

And I'm really was confused with original version of bug:
https://drive.google.com/file/d/1COvG_VyOooe_fyu0p3DVbMZj8uMrnFL9/view
In log at picture, after every line with "Rendering [...].." next line is "Font:NN ..." , where NN is ID in array of SDL_RWops, which used to render this line.

On 2019-03-01 07:09:11 +0000, ace wrote:

I came to the conclusion that using SDL_RWseek(RWops, 0, RW_SEEK_SET) inside TTF_OpenFontIndexRW or TTF_OpenFontRW was bad idea,
so i created feature request in SDL2 tracker to replace macros with functions:
https://bugzilla.libsdl.org/show_bug.cgi?id=4526

On 2019-03-01 08:25:49 +0000, Sylvain wrote:

I also think we shouldn't reset the RWops in SDL_ttf, because user can provide a concatenated ttf files. So I marked this as resolved.

But I haven't understand the issue of your last link "original version of bug".
If there is still a issue or not. So re-opened if this is a mistake.

On 2019-03-01 09:22:57 +0000, ace wrote:

I thought this is a bug, because TTF_OpenFontIndexRW and TTF_OpenFontRW don't return NULL and generate error, if I give them RWops with position at the end of file.

About "original version of bug":
At this moment i am unable call SDL_RWseek(RWops, 0, RW_SEEK_SET) in my C# project.

I use SDL_RWFromMem to create RWops, and i wondered why it works in this way?
You can see that lines with style = 2,6,10,14 used same RWops(18) for TTF_OpenFontIndexRW, but bugged rendering only with style = 6

Or for example lines with style = 0, 4, 8, 12 - used same RWops(19) - but rendered without any problems.

On 2019-03-01 09:40:36 +0000, ace wrote:

OR you can add make small change:
if (position at the еnd of RWops) SDL_RWseek(RWops, 0, RW_SEEK_SET)

It will be solve my problem and don't broken projects, which use this undetermined behavior.

On 2019-03-01 13:47:16 +0000, Sylvain wrote:

Not sure if you can always use SEEK_END eg detect the position at the end of rwops.
For instance RWops could be a http stream (this extension: https://github.com/mgerhardy/SDL_rwhttp )

I look more in details. When it fails, it both:

• returns 0 size metrics (except the advance which is correct)
• returns 0 size after FT rendering.

Those are valid datas .. (like a space)

So for some reason, FreeType seems robust ...
(maybe there would things to report to FT ... http://git.savannah.gnu.org/cgit/freetype/ )

What you can do maybe, is getting a new RWops from the same memory ?

On 2019-03-01 17:32:21 +0000, ace wrote:

What you can do maybe, is getting a new RWops from the same memory ?
I implemented this already, but there is still a small memory leak - I couldn't call SDL_RWclose() cause it's macros :(

I closing this issue: will be wait https://bugzilla.libsdl.org/show_bug.cgi?id=4526
... or maybe make patch by myself (i need SDL 2.0.8, but if this feature request will be accepted, i think it will be only in next release).

Thanks for your help!)

On 2019-03-01 17:56:11 +0000, ace wrote:

I just realize that i could use freesrc=1, when calling TTF_OpenFontIndexRW, and this will be solve my problems at all.

Thanks again!