Bug#1057844: bookworm live images: d-i speech installation not loading sof firmware

[Steve McIntyre]

Package: cdimage.debian.org
Severity: important
Tags: a11y

Testing the gnome live image for the 12.3 release...

Running d-i from that image complained early on about missing Intel SOF
firmware. Later on, the same image finds and loads intel wifi firmware
just fine. Checking on the image, all the firmware debs are sym-linked
in /firmware just fine but we don't have the extra metadata file
(Contents file and dep11 stuff). Early on, AFAICS this means that the
sof firmware isn't detected and loaded.

Indeed, checking a d-i speech installation from this image, I don't
get audio. Argh.

Checking the same machine (Thinkpad T14s) with the amd64 netinst, it
successfully loads firmware and starts the speech installer.

[Steve McIntyre]

Post by Steve McIntyre
Package: cdimage.debian.org
Severity: important
Tags: a11y
Testing the gnome live image for the 12.3 release...
Running d-i from that image complained early on about missing Intel SOF
firmware. Later on, the same image finds and loads intel wifi firmware
just fine. Checking on the image, all the firmware debs are sym-linked
in /firmware just fine but we don't have the extra metadata file
(Contents file and dep11 stuff). Early on, AFAICS this means that the
sof firmware isn't detected and loaded.
Indeed, checking a d-i speech installation from this image, I don't
get audio. Argh.
Checking the same machine (Thinkpad T14s) with the amd64 netinst, it
successfully loads firmware and starts the speech installer.

This isn't new for Bookworm 12.3, is it? There has been no change in the
images generated by live-build regarding the /firmware folder, they have
always contained just the symlinked .deb files.

Nod; this is just the first time I've realised this behaviour.

I only recently started to look deeper into the special folders for the
installer and noticed that the content of /firmware differs between the
netinst and the live images.
Let's discuss this under less time pressure (today is the 12.3 release day).
I think that this ticket could be re-assigned (or cloned) to live-build for
the implementation of the fix.
If needed, could a 12.3.1 live image be released, with just this fix, instead
of waiting for 12.4?

For other reasons, we're about to do a quick-turnaround 12.4. But
otherwise the sentiment is fine, yes! :-)