Modularization Problem with Current CVS?

[Gregory Nowak]

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

Yes, I have to confirm that there does seem to be a modularization
problem, in some configurations at least. I checked this on 2 systems,
and that is why I said some configurations, and not all
configurations.

One system is running slackware 10.1, with gcc 3.3.4. It has
speakup_main built into the kernel, and dtlk built also into the
kernel, which is also the default synth on startup. It has bns as
a module as well.

If I'm running with the dtlk, and attempt to switch to the bns by
doing

echo bns >/proc/speakup/synth_name

, I get a kernel panic. If I'm running with dtlk, and attempt to
switch to the bns by doing

modprobe speakup_bns

, modprobe hangs as Luke describes. If I'm running with dtlk, and
attempt to switch to none by doing

echo none >/proc/speakup/synth_name

, I again get a kernel panic. Switching to dtlk via the echo method,
if running none or bns, seems to work just fine. The only way for me then to use the bns
on this system is to reboot, supply speakup_synth=none as a kernel
parameter, and then echo bns >/proc/speakup/synth_name, or modprobe
speakup_bns once I'm logged in.

The second system I tried this with is a debian testing system, with
gcc 4.1.2, with speakup built into the kernel, bns in the kernel, and
sftsyn as a module. On this machine, I can switch between synths
just fine, regardless of with which synth I boot (bns or none,
followed by loading sftsyn once booted), and I can switch between
synths with echoing the name to speakup_synth, or modprobing the
sftsyn module, regardless of if I do so by switching to none first, or
directly from one synth to the other.

Both machines are running 2.6.18, with cvs speakup checked out as of
Aug. 15th. Specifically

Speakup v-2.00 CVS: Tue Aug 1 11:16:55 EDT 2006

. I'm not sure if the issue lies in the difference between gcc
versions, or if it lies somewhere in the dtlk driver, the bns driver,
the none driver, or in all 3.

Greg


On Mon, Oct 02, 2006 at 02:53:58AM -0500, Luke Davis wrote:
> Is this a new bug, a known bug, or did I just do it wrong?
> 
> I compiled CVS Speakup as modules (all of it).
> 
> After 2.6.18 successfully booted, I followed the instructions in section 
> seven of the spkguide.txt file.
> In so doing, I ran:
> 
> modprobe speakup_dectlk
> 
> At which point modprobe hangs indefinitely.  When I tried it in 
> /etc/modules (speakup_dectlk), which I stupidly did first, I was forced to 
> boot this Ubuntu system with the talking Debian Installation CD, in order 
> to fix /etc/modules.
> 
> When I tried "modprobe speakup_dectlk" from a command line on a console, 
> and it obviously hung, I was able to log into a second console, and do a 
> "killall modprobe", to get it to stop.
> 
> I reconfigured the *exact* same kernel, compiling speakup main directly 
> in, and leaving the synth as a module.
> When I next booted, I was able to run "echo dectlk > 
> /proc/speakup/synth_name", and have the system talking just fine!  (Well, 
> except for the zero followed by alpha problem, which I explained in 
> another message)
> 
> So: what is the problem with having speakup_main as a module, if in fact 
> that is the problem?
> 
> AMD K7 system; kernel 2.6.18; GCC 4.0.3.
> 
> Thanks
> 
> Luke
> 
> _______________________________________________
> Speakup mailing list
> Speakup at braille.uwo.ca
> http://speech.braille.uwo.ca/mailman/listinfo/speakup

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

- --
Free domains: http://www.eu.org/ or mail dns-manager at EU.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFFIYDN7s9z/XlyUyARArHlAKDEdK2nuorUxWBB6hqg8NPotzO7oACdErtO
ct2xOWS9/ohwojrdXZlAix4=
=ytK/
-----END PGP SIGNATURE-----