Speakup Issues

Speakup Issues

[William Hubbs]

On Fri, Mar 29, 2013 at 12:11:33PM -0600, William F. Acker WB2FLW +1-303-722-7209 wrote:
> Someone is probably going to launch an ICBM at my house for asking this 
> question, but, I wonder if we could ditch serial support. Maybe that could 
> be run in user space, and the rest could go into mainline. Do you know if 
> it's just the serial stuff they have a problem with?

Well, the serial code definitely has issues.

For those who don't know, the issue with speakup's serial code is that
we bypass the kernel's serial port drivers and write directly to the
hardware serial ports.

This is not good because we have to hard code the ports to write to into
the speakup code. Also it means that our code doesn't work with all
types of serial ports that the kernel supports.

In order to fix this, it would take working with the serial driver
maintainer in the kernel, at least, and a major rewrite of speakup's
serial code.

The fastest way to take care of this probably would be to write a daemon
which runs in user space. This daemon would act like espeakup, in the
sense that it would read the /dev/softsynth device, but it would send
output to  a serial (or even usb) device that the kernel knows about, e.g.
/dev/ttyS0.

I have been working on a usb device driver for the tripple talk for a while;
I should get back to finishing that.

The disadvantage of this of course is that hardware synthesizers then
would be subject to the same limitation as the software synthesizer; you
would have to wait until some point in the boot process when a daemon
could be started before speech would start. I think hardware speech
could start up pretty early; probably around the same time as brltty.

another issue right now is that when data is sent to the software
synthesizer, on an smp machine, data gets randomly dropped. If you
review the screen everything is there as it should be, but when the data
is initially sent to the synthesizer, parts of it do not get read.

This hasn't been fixed, because it is extremely difficult to debug and
reproduce.

Another question is that we still have drivers for old internal ISA
synthesizers, such as the doubletalk pc, dectalk pc, accent pc and
keynote pc. Are these being used? is it even possible to find a machine
any more that supports ISA slots?

William

Re: Speakup Issues

[Adam Myrow]

Regarding the problems with serial ports, why can't Speakup tie into the 
serial console code?  There is already code to make a serial port act as 
the console, so I would think that Speakup could use some of that code. 
If the daemon route is chosen instead, could the daemon be run from an 
initial RAM disk?  That would get it to start up before the root 
filesystem mounts, and that is when most problems happen, if they are 
going to.  The whole point of having Speakup support serial synthesizers 
from boot is to deal with such problems as the root filesystem refusing to 
mount, and being able to read the error message to see what is wrong.

Re: Speakup Issues

[Jason White]

Adam Myrow  <speakup@linux-speakup.org> wrote:
>Regarding the problems with serial ports, why can't Speakup tie into the 
>serial console code?  There is already code to make a serial port act as 
>the console, so I would think that Speakup could use some of that code. 

I think the underlying problem is finding people with kernel development
experience who are willing to work on the problem. With that solved, I suspect
the rest of the issues would go away with the right technical cooperation and
liaison with the kernel community.

Re: Speakup Issues

[Gregory Nowak]

On Fri, Mar 29, 2013 at 04:59:24PM -0500, William Hubbs wrote:
> Another question is that we still have drivers for old internal ISA
> synthesizers, such as the doubletalk pc, dectalk pc, accent pc and
> keynote pc. Are these being used?

Yes, I still have an old box here with an isa slot with a doubletalk
pc. I mainly use it these days to access other linux boxes, because
I'm very fond of my doubletalk, and don't want to give it up if I
don't have to. I have nothing against espeak/espeakup, it's just that
hearing that "doubletalk pc found" message invariably brings a smile to my
face. Also, maybe it's my imagination, but that doubletalk pc on old
hardware seems to still be slightly more responsive than
espeak/espeakup on modern hardware. I have also idly toyed with the
idea of getting one of these:

http://www.arstech.com/item--usb2isa.html

and found myself wondering if speakup could be made to drive a doubletalk pc
through this converter.

Greg


-- 
web site: http://www.gregn.net
gpg public key: http://www.gregn.net/pubkey.asc
skype: gregn1
(authorization required, add me to your contacts list first)

--
Free domains: http://www.eu.org/ or mail dns-manager@EU.org

Re: Speakup Issues

[Steve Holmes]

I would say "Ditch the old ISA support." Sorry Greg:).  I say this for
the old simple reason that ISA buses aren't even made anymore AFAIK.
I realize there may still be some older legacy hardware out there but
after some point, it really isn't practical to support it anymore.
People have made similar remarks about serial ports in general but I
have a feeling that serial ports are living a longer life than ISA
buses and if this support is keeping Speakup out of the Kernel
spotlight then we should dump it.  

Greg, I have a USB Tripple Talk USB here and would love to be able to
use it with Speakup and I think it would be quite responsive provided
the USB synth support is good and tight.

On Fri, Mar 29, 2013 at 06:32:51PM -0700, Gregory Nowak wrote:
> On Fri, Mar 29, 2013 at 04:59:24PM -0500, William Hubbs wrote:
> > Another question is that we still have drivers for old internal ISA
> > synthesizers, such as the doubletalk pc, dectalk pc, accent pc and
> > keynote pc. Are these being used?
> 
> Yes, I still have an old box here with an isa slot with a doubletalk
> pc. I mainly use it these days to access other linux boxes, because
> I'm very fond of my doubletalk, and don't want to give it up if I
> don't have to. I have nothing against espeak/espeakup, it's just that
> hearing that "doubletalk pc found" message invariably brings a smile to my
> face. Also, maybe it's my imagination, but that doubletalk pc on old
> hardware seems to still be slightly more responsive than
> espeak/espeakup on modern hardware. I have also idly toyed with the
> idea of getting one of these:
> 
> http://www.arstech.com/item--usb2isa.html
> 
> and found myself wondering if speakup could be made to drive a doubletalk pc
> through this converter.
> 
> Greg
> 
> 
> -- 
> web site: http://www.gregn.net
> gpg public key: http://www.gregn.net/pubkey.asc
> skype: gregn1
> (authorization required, add me to your contacts list first)
> 
> --
> Free domains: http://www.eu.org/ or mail dns-manager@EU.org
> _______________________________________________
> Speakup mailing list
> Speakup@linux-speakup.org
> http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup

Re: Speakup Issues

[John G. Heim]

How does a serial console work? If you can configure a serial console, why can't you configure a serial speech synthesizer? Or does the kernel team have problems with serial consoles too?

How does a serial terminal work?


On Mar 29, 2013, at 4:59 PM, William Hubbs wrote:

> On Fri, Mar 29, 2013 at 12:11:33PM -0600, William F. Acker WB2FLW +1-303-722-7209 wrote:
>> Someone is probably going to launch an ICBM at my house for asking this 
>> question, but, I wonder if we could ditch serial support. Maybe that could 
>> be run in user space, and the rest could go into mainline. Do you know if 
>> it's just the serial stuff they have a problem with?
> 
> Well, the serial code definitely has issues.
> 
> For those who don't know, the issue with speakup's serial code is that
> we bypass the kernel's serial port drivers and write directly to the
> hardware serial ports.
> 
> This is not good because we have to hard code the ports to write to into
> the speakup code. Also it means that our code doesn't work with all
> types of serial ports that the kernel supports.
> 
> In order to fix this, it would take working with the serial driver
> maintainer in the kernel, at least, and a major rewrite of speakup's
> serial code.
> 
> The fastest way to take care of this probably would be to write a daemon
> which runs in user space. This daemon would act like espeakup, in the
> sense that it would read the /dev/softsynth device, but it would send
> output to  a serial (or even usb) device that the kernel knows about, e.g.
> /dev/ttyS0.
> 
> I have been working on a usb device driver for the tripple talk for a while;
> I should get back to finishing that.
> 
> The disadvantage of this of course is that hardware synthesizers then
> would be subject to the same limitation as the software synthesizer; you
> would have to wait until some point in the boot process when a daemon
> could be started before speech would start. I think hardware speech
> could start up pretty early; probably around the same time as brltty.
> 
> another issue right now is that when data is sent to the software
> synthesizer, on an smp machine, data gets randomly dropped. If you
> review the screen everything is there as it should be, but when the data
> is initially sent to the synthesizer, parts of it do not get read.
> 
> This hasn't been fixed, because it is extremely difficult to debug and
> reproduce.
> 
> Another question is that we still have drivers for old internal ISA
> synthesizers, such as the doubletalk pc, dectalk pc, accent pc and
> keynote pc. Are these being used? is it even possible to find a machine
> any more that supports ISA slots?
> 
> William
> 
> _______________________________________________
> Speakup mailing list
> Speakup@linux-speakup.org
> http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup
> 

Re: Speakup Issues

[Gregory Nowak]

On Sat, Mar 30, 2013 at 05:12:55AM -0700, Steve Holmes wrote:
> I would say "Ditch the old ISA support." Sorry Greg:).  I say this for
> the old simple reason that ISA buses aren't even made anymore AFAIK.

Ok, I decided to do a search to see if this is true. Yes, the ISA bus
still exists in today's pc's, but I was wondering if it's still
possible to get motherboards with an isa slot:

<http://www.google.com/search?ie=ISO-8859-1&hl=en&source=hp&q=motherboard%2Bwith%2Bisa%2Bslot&btnG=Google+Search&gbv=1>

Granted, the few I quickly checked out from these results seem to cost
quite the $$$, so it would make more sense to get a new usb
speakup-supported synth, which would probably cost the same, if not
less. The point is though that motherboards with isa slots seem to
still be available if someone wants one badly enough, and can afford
it. Also, while most of what I glanced through is designed for older
processors, at least one of the boards I noticed supports core i5/i7
processors.

> I realize there may still be some older legacy hardware out there but
> after some point, it really isn't practical to support it anymore.
> People have made similar remarks about serial ports in general but I
> have a feeling that serial ports are living a longer life than ISA
> buses and if this support is keeping Speakup out of the Kernel
> spotlight then we should dump it.  

I agree. However, no one has said so far as far as I know that support
for isa synths is keeping speakup out of the kernel spotlight. William
simply asked if anybody is still using these, and my answer is yes.

> 
> Greg, I have a USB Tripple Talk USB here and would love to be able to
> use it with Speakup and I think it would be quite responsive provided
> the USB synth support is good and tight.
> 

Agreed again. I do seem to recall though that attempts were made way
back when to get the manufacturer of the tripletalk to disclose the
specs, and that those attempts failed. I remember telling myself
back then that I wouldn't give my business to a company which refused
to make its hardware open. Has this changed, or would tripletalk
support as of today need to be hacked together? If this hasn't
changed, and won't have changed by the time speakup supports the
tripletalk, then I guess the question for me would be first, can I
afford to get one at that time, and second, do I want one badly enough
to go back on my principals? Not to hijack the thread, but it would be
interesting to get viewpoints as far as giving business to companies
which refuse to open their hardware goes.

Greg


-- 
web site: http://www.gregn.net
gpg public key: http://www.gregn.net/pubkey.asc
skype: gregn1
(authorization required, add me to your contacts list first)

--
Free domains: http://www.eu.org/ or mail dns-manager@EU.org

Re: Speakup Issues

[Jason White]

Gregory Nowak  <speakup@linux-speakup.org> wrote:

>I agree. However, no one has said so far as far as I know that support
>for isa synths is keeping speakup out of the kernel spotlight. William
>simply asked if anybody is still using these, and my answer is yes.

If it isn't creating a significant maintenance burden, I would recommend
leaving the code in place.

As an aside, I have a DECTALK-PC card here that requires an ISA slot, but no
machine in which to run it and no current need for it. At some point I'll
donate it to a worthy recipient, who will then have the interesting experience
of discovering whether or not it works anymore.

Re: Speakup Issues

[Steve Holmes]

Yeah, I agree with your views about supporting a company that hides
its specs.  It just happened that I bought this synth from a private
party as a used item.  But the principle is still true though.

On Sat, Mar 30, 2013 at 05:07:27PM -0700, Gregory Nowak wrote:
> Agreed again. I do seem to recall though that attempts were made way
> back when to get the manufacturer of the tripletalk to disclose the
> specs, and that those attempts failed. I remember telling myself
> back then that I wouldn't give my business to a company which refused
> to make its hardware open. Has this changed, or would tripletalk
> support as of today need to be hacked together? If this hasn't
> changed, and won't have changed by the time speakup supports the
> tripletalk, then I guess the question for me would be first, can I
> afford to get one at that time, and second, do I want one badly enough
> to go back on my principals? Not to hijack the thread, but it would be
> interesting to get viewpoints as far as giving business to companies
> which refuse to open their hardware goes.
> 
> Greg
> 
> 
> -- 
> web site: http://www.gregn.net
> gpg public key: http://www.gregn.net/pubkey.asc
> skype: gregn1
> (authorization required, add me to your contacts list first)
> 
> --
> Free domains: http://www.eu.org/ or mail dns-manager@EU.org
> _______________________________________________
> Speakup mailing list
> Speakup@linux-speakup.org
> http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup

Re: Speakup Issues

[William Hubbs]

On Sat, Mar 30, 2013 at 04:34:23PM -0500, John G. Heim wrote:
> How does a serial console work? If you can configure a serial console,
> why can't you configure a serial speech synthesizer? Or does the kernel
> team have problems with serial consoles too?

I haven't looked at this code in a long time, but I do remember that it
is unsuitable for speakup.

> How does a serial terminal work?

Serial terminals are completely different; they are handled in user
space, by getty, much later in the boot process.

William

Re: Speakup Issues

[Janina Sajka]

Ah, Greg, I'm jealous!


But, I may catch up to you. I do have a pristine, never used, Socket 7
mobo,  appropriate CPU and RAM here. Just need to find a case and psu
for it, I suppose.  I've got about 3 ISA Doubletalk cards.

Janina

Gregory Nowak writes:
> On Fri, Mar 29, 2013 at 04:59:24PM -0500, William Hubbs wrote:
> > Another question is that we still have drivers for old internal ISA
> > synthesizers, such as the doubletalk pc, dectalk pc, accent pc and
> > keynote pc. Are these being used?
> 
> Yes, I still have an old box here with an isa slot with a doubletalk
> pc. I mainly use it these days to access other linux boxes, because
> I'm very fond of my doubletalk, and don't want to give it up if I
> don't have to. I have nothing against espeak/espeakup, it's just that
> hearing that "doubletalk pc found" message invariably brings a smile to my
> face. Also, maybe it's my imagination, but that doubletalk pc on old
> hardware seems to still be slightly more responsive than
> espeak/espeakup on modern hardware. I have also idly toyed with the
> idea of getting one of these:
> 
> http://www.arstech.com/item--usb2isa.html
> 
> and found myself wondering if speakup could be made to drive a doubletalk pc
> through this converter.
> 
> Greg
> 
> 
> -- 
> web site: http://www.gregn.net
> gpg public key: http://www.gregn.net/pubkey.asc
> skype: gregn1
> (authorization required, add me to your contacts list first)
> 
> --
> Free domains: http://www.eu.org/ or mail dns-manager@EU.org
> _______________________________________________
> Speakup mailing list
> Speakup@linux-speakup.org
> http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup

-- 

Janina Sajka,	Phone:	+1.443.300.2200
			sip:janina@asterisk.rednote.net
		Email:	janina@rednote.net

Linux Foundation Fellow
Executive Chair, Accessibility Workgroup:	http://a11y.org

The World Wide Web Consortium (W3C), Web Accessibility Initiative (WAI)
Chair,	Protocols & Formats	http://www.w3.org/wai/pf
	Indie UI			http://www.w3.org/WAI/IndieUI/

Re: Speakup Issues

[Jason White]

Janina Sajka <janina@rednote.net> wrote:
> But, I may catch up to you. I do have a pristine, never used, Socket 7 mobo,
> appropriate CPU and RAM here. Just need to find a case and psu for it, I
> suppose.  I've got about 3 ISA Doubletalk cards.

If you also want an ISA DECTALK-PC card in unknown condition, I can supply. It
was taken from a disused machine when the university sent the latter to
component recycling, leaving me with the card.

These items do exist, which is why leaving the drivers in the kernel would be
a good idea, unless of course they become a maintenance burden.

Re: speakup issues

[Steve Holmes]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: RIPEMD160

As far as other hardware synths go, I use a Speakout from time to time
and have not experienced any issues at all with the thing.  It just
responds as good as ever.

On Fri, May 29, 2009 at 09:20:00AM -0400, Kirk Reiser wrote:
> As far as I've ever been able to determine that stuttering is common
> to all RC Systems synths based on the DoubleTalk chip set.  It seems
> to be set off by something I've never fully nailed down.  I think
> dashes before some strings will do it but I'm not really sure what
> else starts the stuttering.
>
>   Kirk
> --
> Kirk Reiser				The Computer Braille Facility
> e-mail: kirk@braille.uwo.ca		University of Western Ontario
> phone: (519) 661-3061
> _______________________________________________
> Speakup mailing list
> Speakup@braille.uwo.ca
> http://speech.braille.uwo.ca/mailman/listinfo/speakup
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEUEAREDAAYFAko2PE8ACgkQWSjv55S0LfHhqwCVFUjkyREtuFMqqOlt0Wv5JK8U
5ACguvCeF7f9JLjC3FDDo/lvaWJhDrg=
=avYj
-----END PGP SIGNATURE-----

Re: speakup issues

[Kirk Reiser]

As far as I've ever been able to determine that stuttering is common
to all RC Systems synths based on the DoubleTalk chip set.  It seems
to be set off by something I've never fully nailed down.  I think
dashes before some strings will do it but I'm not really sure what
else starts the stuttering.

   Kirk
--
Kirk Reiser				The Computer Braille Facility
e-mail: kirk@braille.uwo.ca		University of Western Ontario
phone: (519) 661-3061

Re: speakup issues

[Tony Baechler]

Hi,

I did notice an issue with the Trippletalk not reported here, but I 
think it's unique to that synth.  Yes, there is definitely a delay when 
using the keypad plus before speech starts, either with the Doubletalk 
LT or Trippletalk.  If using the regular reading keys, there is no 
delay.  I don't recall noticing that with my DEC Express but it's been 
several months since I've used Speakup with that synthesizer.  In my 
brief tests with software speech, I didn't notice this and actually 
Speakup seemed far more responsive than usual.

The problem with the Trippletalk that I would like to report is a bad 
case of stuttering, but only sometimes.  I think it's unique to that 
synth because the Doubletalk LT didn't do it, but it can generally be 
repeated by using keypad plus.  It used to happen sometimes in Windows 
with the same synthesizer but not all the time.  Again, it only happens 
when I read the entire screen and only sometimes.  I can't think of a 
case where it can be consistently reproduced, but one thing that always 
seems to do it is the word "file" near the top of the screen, such as 
the "File" menu in many Windows apps or if the word shows up at the top 
of the screen with Speakup.  It will stutter and repeat a few times 
before reading the rest of the screen normally.

Michael, does the Jupiter screen reader support 2.6.29 kernels with 
SMP?  If it supports recent kernels on the x86-64 arch and doesn't 
require a custom kernel to be compiled, I would like to try it.  I had 
no luck with yasr at all and the manual leaves much to be desired.  Is 
it in Debian or where can it be downloaded?

Re: speakup issues

[Samuel Thibault]

William Hubbs, le Thu 28 May 2009 09:15:07 -0500, a écrit :
> To make way for software voice
> synthesis, SpeakUP has become a piece of junk for everything else.

The changes that have been done were made to get speakup work with
recent kernels and things like preemption, SMP, etc.  Without that you
wouldn't have support _at all_.  Software support has nothing to do with
that.

Samuel

speakup issues

[William Hubbs]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

All,

This is part of a message which was posted to the gentoo-accessibility
list.  I am bringing it here because it is about speakup, and I will do
my best to answer it here.

The first thing I want to respond to is the allegation that speakup's
hardware speech synth support is "falling apart" to make way for
software speech.  We are doing our best to support both.  We have added
a way to support software speech, but that should not affect hardware
speech at all.  If you have issues report them here and we will do our
best to fix them.  Or, better yet, submit a patch and we will look it
over and commit it if it fixes the issue.  Suggestions, bug reports, and
patches are always welcome.

Regarding Janina's trippletalk not working:  There is not enough
information for us yet to determine what is happening.  I know I use a
trippletalk with the latest speakup and it works.  Also, there is
another person who has replied to the thread and uses a trippletalk
without issues.  I also believe Kirk uses one.

You said that speakup crashes when you log onto irc.  What do you mean
by "crashes".  Locks up?  Does it lock up the box to the point that you
have to reboot?  Are you able to recover some how?  How do you get
speech back?  Do you see any messages in syslog?

I do see the issue you are talking about with the ltlk taking a second
to start reading the screen when the keypad plus key is pressed.  Has
anyone else on the list seen this with other synths?

Michael, it would be helpful if you would work with us on this list so
that we can find out what is causing your issues and get them fixed.

Thanks,

William

- ----- Forwarded message from Gaijin <gaijin@clearwire.net> -----

	I'm not talking about that aspect.  I know SpeakUP can be
compiled in as integrated modules.  I use the 1.1 version of  the ltlk,
2.00 version of speakup in Shane's Debian Etch kernel, because it's the
version that crashes the least and works the fastest.  The latest
SpeakUP builds using the ltlk module take a full second or more to begin
reading the screen after pressing keypad plus.  It is nice having a
hardware synth available to correct problems before the software synth
boots, but from all I've seen, the hardware synth modules are falling
apart.  Speakup crashes every time I log onto irc.freenode.net using my
hardware synthesizer, Janina's Tripletalk over on the SpeakUP mailing
list has stopped working, etc.  To make way for software voice
synthesis, SpeakUP has become a piece of junk for everything else.
	Karl Dahlk's Jupiter screen reader can send plaintext to the
synthesizers that support that mode of operation, so I plan on compiling
in both of them and seeing which is best for the console.

			Michael



- ----- End forwarded message -----

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.11 (GNU/Linux)

iEYEARECAAYFAkoenGsACgkQblQW9DDEZTgO1gCfQ1CgcCCkUm6FQI0jS43L59x8
rckAoIeMXIzuaejUKnAQ+kNJS/QgSMRK
=xBRz
-----END PGP SIGNATURE-----