[vlc-devel] commit: Do not ignore teletext initial page (close #1997). (Laurent Aimar )

Derk-Jan Hartman hartman at videolan.org
Mon Sep 8 21:12:34 CEST 2008


On 8 sep 2008, at 19:54, Marian Ďurkovič wrote:
> Sure - please fetch http://md.bts.sk/telx/ct1.ts
>
> Select teletext page 898 - the test picture.
> It should look like this:
>
> http://www.ceskatelevize.cz/teletext/index.php?page=898&ver=&channel=
>
> The capital letters are (mostly) OK, only "Ť" is missing.
> However the small special characters at rows 08 and 09 are completely
> wrong.
>
> Those characters are transmitted differently - the capitals which  
> require
> advanced teletext capabilities work almost fine.

Looks like a bug in zvbi to me.

If I look at the zvbi headers, I find
{ 0x0023, 0x016F, 0x010D, 0x0165, 0x017E, 0x00FD, 0x00ED, 0x0159,  
0x00E9, 0x00E1, 0x011B, 0x00FA, 0x0161 },

Compared to telx.c
{ 0x0023, 0x016f, 0x010d, 0x0165, 0x017e, 0x00fd, 0x00ed, 0x0159,  
0x00e9, 0x00e1, 0x0115, 0x00fa, 0x0161 }

That means that the 2nd to last char of the table differs.
According to table 36 of ETSI EN 300 706, that would be an e with  
accent with an inverted accent circonflexe.
However it seems many more characters than this one are affected. So  
looking at the values of the characters and comparing them to the  
tables it seems as if zvbi thinks this is Turkish instead of Czech/ 
Slovak.

Now if we take a look at table 32 of the TTXT spec, then we see that  
in the third triplet configuration for selection bits, turkish  
occupies the same control bits row as czech/slovak does for the first  
2 selection bit patterns.

That means that something goes wrong in character_set_designation of  
zvbi's teletext.c file. Either the TTXT source is incorrect, or the  
way that zvbi detects the charset fails.

That's about as far as I feel like debugging this :D

DJ





More information about the vlc-devel mailing list