On Fri, 2017-03-31 at 08:15 +0300, Paul Fertser wrote: > On Thu, Mar 30, 2017 at 10:04:24PM +0100, Ben Hutchings wrote: > > On Thu, 2017-03-30 at 09:22 +0800, Paul Wise wrote: > > > Source: linux > > > Version: 4.10~rc6-1~exp1 > > > Severity: wishlist > > > X-Debbugs-CC: open-ath9k-htc-firmware@packages.debian.org > > > > > > Now that open-ath9k-htc-firmware has been accepted into Debian > > > unstable, please add "Recommends: firmware-ath9k-htc" to the > > > metadata for the linux-image-* packages in Debian experimental. > > Not many linux-image-* users have ath9k-htc hardware so I do not see > how this recommendation can make sense here. This is also true for most of the devices supported by firmware-linux- free, but it's small so it shouldn't hurt. > The package should have provided appropriate AppStream metainformation > so Debian should be able to suggest installing it when the device is > plugged in for the first time. Unfortunately I don't think we have all the infrastructure in place for that yet. > > As this firmware has gone through at least one ABI bump, I think we > > need to plan for a future ABI bump. > > So far the idea was to upload a package named firmware-ath9k-htc-1.5.0 > after the next ABI bump. There's no reason why > firmware-ath9k-htc-1.5.0 shouldn't be able to co-exist on the same > system with e.g. firmware-ath9k-htc-1.6.0, as the user should be able > to choose different kernel versions on boot, and hence different > firmware versions will be appropriate. > > > Therefore: > > - You should not name the files as simply '1.dev.0' versions, but by > > the implemented ABI version (as the driver expects by default). > > The code that's currently packaged is definitely not 1.4.0 code, it > got some non-trivial changes (not affecting ABI though) after the > 1.4.0 was released. So naming an intermediate version in any way other > than 1.dev.0 would only add to the confusion IMHO. So install your files with the real version number and make a symlink with the '1.4.0' name. Ben. > Probably it would make sense to have the minor number indicate a > subversion of same-ABI firmwares, but for some reasons the kernel > driver maintainers decided against that. > > I hope Oleksij will correct me if I'm missing something here. > -- Ben Hutchings To err is human; to really foul things up requires a computer.
Attachment:
signature.asc
Description: This is a digitally signed message part