From: Dominik Riebeling <dominik.riebeling_at_gmail.com>
Date: Mon, 20 Feb 2012 23:58:17 +0100

I've seen quite a bit of false information about the state for
voicefile generation with Rockbox Utility on this list lately. I've
unsubscribed from this list years ago and resubscribed to write this
message so I unfortunately can't reply to the appropriate thread.
Since I've left because of the attitude of the people on this list I
might leave again soon.

But for now let me explain the current state: right now people are
mixing up several independent problems. Please read the *whole*
explanation before complaining.

Rockbox Utility 1.2.7 introduced multithreading for voicefile
generation. While this increased generation speed notably on multicore
machines it also introduced a problem for some people. However, this
problem (symptom: generation stopping at some point complaining about
a missing file) only shows up in some special cases. To get affected
you need to have ALL of the following conditions met: (a) you're on
Windows, (b) you're using a multicore machine, (c) you're using SAPI
to generate the voicefile *and* (d) you use a voice that is affected
-- it seems to affect commercial voices pretty often, the voices
shipped with Windows (MS Sam and Anna) do not show the problem. A
workaround for this was added with Rockbox Utility 1.2.9, which
unfortunately turned out to not fix the problem completely. The issue
has been fixed with Rockbox Utility 1.2.12. Either way, simply
claiming that voicefile generation didn't work with Rockbox Utility
after 1.2.6 is completely nonsense. Rockbox Utility on Linux and Mac
were never affected by this issue.

Rockbox itself recently switched its version control system to git.
This unfortunately broke some of the infrastructure on the server.
Rockbox Utility needs to ask the server about the voice strings -- the
language files on the player only carry information about the strings
to display to keep their size down. When querying the server for the
voice strings it is necessary to tell the server the installed Rockbox
version. Since git has a different concept of naming versions but the
script generating the list of strings hasn't been updated the new
version numbers aren't recognized yet. As a result you'll get a
"downloaded file was empty" error message. This will *only* occur if
you are using a build that is (a) not a release up to and including
3.10 and (b) a build that has been made after the switch to git (the
latest build based on svn is r31646-20120109). This issue it *not*
related to the multithreading issue described above in any way. A
temporary solution for the server-side issue is currently in review
and hopefully will be available soon.

If you want to report problems do *not* report it on this list. Most
Rockbox developers have left this list because of the attitude of its
subscribers (like I did). Nobody working on Rockbox Utility is
actually reading this list. Do *not* "IIRC" that people are reading
this list. If you want to report problems use the proper tool -- the
bugtracker. If you don't know where to find that just ask Google:
http://lmgtfy.com/?q=rockbox+bugtracker

Please stop spreading wrong information. If you make claims please
provide proof to backup your claim. You're saving others lots of hours
by not making them follow wrong directions. Thanks.


 - Dominik
-------------------------------------------------------------------
Unsubscribe: http://cool.haxx.se/cgi-bin/mailman/listinfo/rockbox
FAQ: http://www.rockbox.org/twiki/bin/view/Main/GeneralFAQ
Etiquette: http://www.rockbox.org/mail/etiquette.html
Received on 2012-02-21