[Sugar-devel] Problem with gstreamer espeak plugin

[James Simmons]

Aleksey,

I recently tried to use text to speech on a French file with Read
Etexts and it behave really strangely.  When speaking a page the
highlight jumps ahead of the word being spoken.  Speech works fine,
it's just the highlight that gets messed up.  It does this on most
pages, so I'll give one example.  On page three or so the value for
the tag name goes up to 57, then skips to 145:

word_count 53
word_count 53
word_count 54
word_count 54
word_count 54
word_count 55
word_count 55
word_count 56
word_count 56
word_count 57
word_count 57
word_count 145
word_count 145
word_count 145
word_count 145
word_count 146
word_count 146
word_count 146
word_count 147
word_count 147
word_count 148
word_count 148
word_count 149

The relevant park of the marked up text being sent to espeak looks like this:

<speak> <mark name="0"/>Ce<mark name="1"/>système<mark
name="2"/>orographique<mark name="3"/>appartenait<mark
name="4"/>à<mark name="5"/>la<mark name="6"/>portion<mark
name="7"/>la<mark name="8"/>plus<mark name="9"/>sauvage<mark
name="10"/>de<mark name="11"/>la<mark name="12"/>Transylvanie,<mark
name="13"/>comprise<mark name="14"/>sous<mark name="15"/>la<mark
name="16"/>dénomination<mark name="17"/>de<mark
name="18"/>comitat<mark name="19"/>de<mark name="20"/>Klausenburg<mark
name="21"/>ou<mark name="22"/>Kolosvar.<mark name="23"/>Curieux<mark
name="24"/>fragment<mark name="25"/>de<mark name="26"/>l'empire<mark
name="27"/>d'Autriche,<mark name="28"/>cette<mark
name="29"/>Transylvanie,<mark name="30"/>«<mark
name="31"/>l'Erdely<mark name="32"/>»<mark name="33"/>en<mark
name="34"/>magyar,<mark name="35"/>c'est-à-dire<mark name="36"/>«<mark
name="37"/>le<mark name="38"/>pays<mark name="39"/>des<mark
name="40"/>forêts<mark name="41"/>».<mark name="42"/>Elle<mark
name="43"/>est<mark name="44"/>limitée<mark name="45"/>par<mark
name="46"/>la<mark name="47"/>Hongrie<mark name="48"/>au<mark
name="49"/>nord,<mark name="50"/>la<mark name="51"/>Valachie<mark
name="52"/>au<mark name="53"/>sud,<mark name="54"/>la<mark
name="55"/>Moldavie<mark name="56"/>à<mark name="57"/>l'ouest.<mark
name="58"/>Étendue<mark name="59"/>sur<mark name="60"/>soixante<mark
name="61"/>mille<mark name="62"/>kilomètres<mark
name="63"/>carrés,<mark name="64"/>soit<mark name="65"/>six<mark
name="66"/>millions<mark name="67"/>d'hectares<mark name="68"/>--<mark
name="69"/>à<mark name="70"/>peu<mark name="71"/>près<mark
name="72"/>le<mark name="73"/>neuvième<mark name="74"/>de<mark
name="75"/>la<mark name="76"/>France<mark name="77"/>--,<mark
name="78"/>c'est<mark name="79"/>une<mark name="80"/>sorte<mark
name="81"/>

I'm hoping you or someone else can see a pattern that I'm missing.  If
anyone wants to try to recreate the problem the text is Le chateau de
Carpathia by Jules Verne, UTF-8 from Project Gutenberg.

Thanks,

James Simmons