Questions regarding Debian nvidia packages

Vincent Cheng vcheng at
Fri Sep 12 08:24:25 UTC 2014

Hi Andreas,

On Tue, Sep 9, 2014 at 11:27 PM, Andreas Beckmann <anbe at> wrote:

>> 2) What's the process of splitting and uploading a new legacy driver?
>> According to [3], the 340 series is the next legacy driver (being the
>> last version to support pre-Fermi cards it seems), and I've yet to
>> stumble across documentation describing all the gotchas that are
>> involved in packaging a new legacy driver. As for timing, would you
>> agree that this is best done after Jessie is released?
> That sounds like a good idea -> nothing newer than 340.xx max go to
> unstable, and we have a long-term release in jessie where we may get
> upstream update releases (think security)

Ack. What I really wanted to ask was _how_ to go about preparing a new
legacy package. I suppose most of the process is manual? e.g. updating
nvidia-detect, updating alternatives, updating package names, probably
lots of other stuff I haven't thought of?

>> 3) Upstream bumblebee devs want me to ask you why the nvidia module
>> was renamed (nvidia-current in latest driver, nvidia-xxx for the
>> legacy drivers AFAIK), and frankly I don't really understand the
>> rationale myself. Would you care to explain to upstream on github [4]?
> to be able to have both current and legacy kernel modules available and
> decide at runtime (i.e without installing/removing packages) which one
> to load
> can't use alternatives on the module itself since there may be any
> number of kernels installed


>> Also, they've pointed out that if the nvidia module is renamed, why
>> isn't nvidia-uvm renamed as well?
> so far there was no legacy -uvm I assume that once the 340.xx series is split off into a legacy
package, the uvm module will also be renamed and versioned as well?


More information about the pkg-nvidia-devel mailing list