Sonar GNU/Linux merges with Vinux

Linux for blind general discussion blinux-list at redhat.com
Sat Apr 29 22:26:02 UTC 2017


Luke Yelavich here, reply inline.

On Wed, Apr 26, 2017 at 09:04:56PM AEST, Linux for blind general discussion wrote:
> Am I correct in saying that while in theory it could be done, isn't it
> impractical? Taking a standard initramfs on a standard Debian-based system,
> how would you know what sound drivers to include? I'm sure initramfs-tools
> doesn't do this, even if you include all modules. That would add a huge
> amount of bloat, possibly running you out of memory. You would need at least
> libespeak and espeakup, right?

Well at initramfs build time, you could detect the driver needed for any 
present sound hardware, firstly by looking at /proc/modules, and then also 
using lspci to look up require kernel modules for any audio cards present. As 
for other components, yes libespeak and espeakup, and of course 
speech-dispatcher if you use another synth, but that requires a little more 
work.

Having thought about this though, it may be worth putting together a special 
build of speech-dispatcher that is usable in an initramfs. I may end up doing 
this anyway just for kicks.

> The only way I could see this being done is if you build your own custom
> initramfs for your system and sound card with your sound card drivers. I
> remember there was a lot of debate about this with D-I. The smaller D-I
> images would be too big with the sound card drivers. The kernel can
> autodetect and load the right module, but again, almost all sound modules
> would have to be included, right?

As above, no, all sound modules would not need to be included. When checking 
loaded modules, you could certainly create a file that a script in the 
initramfs could use to load the modules that were present on the installed 
system, such that any sound modules that are not easily auto-detectable get 
loaded.

Luke




More information about the Blinux-list mailing list