Re: [PATCH 09/20] Add os-release file

[Demi Marie Obenour <demiobenour@gmail.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 3453 bytes --]

On 9/8/25 05:12, Alyssa Ross wrote:
> Demi Marie Obenour <demiobenour@gmail.com> writes:
> 
>> systemd-sysupdate expects one to exist and it's a good idea to have one
>> anyway.  Some third-party dependencies might check for it.
>>
>> Signed-off-by: Demi Marie Obenour <demiobenour@gmail.com>
>> ---
>>  host/rootfs/Makefile               |  1 +
>>  host/rootfs/etc/os-release         | 12 ++++++++++++
>>  host/rootfs/etc/os-release.license |  2 ++
>>  img/app/Makefile                   |  1 +
>>  img/app/etc/os-release             | 12 ++++++++++++
>>  img/app/etc/os-release.license     |  2 ++
>>  vm/sys/net/Makefile                |  1 +
>>  vm/sys/net/etc/os-release          | 12 ++++++++++++
>>  vm/sys/net/etc/os-release.license  |  2 ++
>>  9 files changed, 45 insertions(+)
>>
> 
> In general, I wouldn't want most software to be making decisions on
> /etc/os-release.  (systemd-sysupdate is a special case here in having a
> good reason to do it.)  Maybe in img/app we need it for compatibility
> with arbitrary stuff, but if anything in vm/sys/net is looking at
> os-release I'd rather it crash and I find out about it so I could fix
> it.

I don't think it would crash, though, just fall back to other behavior
(which might not be desirable).

> If we do have an /etc/os-release file on the host though, would be great
> if we could re-use it when building the UKI in release/live.
> (dump.erofs is probably useful here.)

That should be doable.

>> diff --git a/host/rootfs/Makefile b/host/rootfs/Makefile
>> index 6cdbac201257faedb70344bcfd5cf9d4fd25b507..4faaccab8cb01d57ef7c48c01eb6fb1326cea4a0 100644
>> --- a/host/rootfs/Makefile
>> +++ b/host/rootfs/Makefile
>> @@ -17,6 +17,7 @@ FILES = \
>>  	etc/mdev/listen \
>>  	etc/mdev/net/add \
>>  	etc/mdev/wait \
>> +	etc/os-release \
>>  	etc/parse-devname \
>>  	etc/passwd \
>>  	etc/s6-linux-init/env/WAYLAND_DISPLAY \
>> diff --git a/host/rootfs/etc/os-release b/host/rootfs/etc/os-release
>> new file mode 100644
>> index 0000000000000000000000000000000000000000..536183411aa94b727f045c4623c29d66503738be
>> --- /dev/null
>> +++ b/host/rootfs/etc/os-release
>> @@ -0,0 +1,12 @@
>> +NAME="Spectrum OS"
> 
> NAME="Spectrum".  There's no "OS" is the name.
> 
>> +ID="spectrum"
>> +PRETTY_NAME="Spectrum OS 0.0.0-alpha0"
>> +VERSION="0.0.0-alpha0"
>> +VERSION_ID="0"
>> +IMAGE_ID="Spectrum-OS-Host"
> 
> The documentation for this field says "A lower-case string".
> 
>> +IMAGE_VERSION="0"
> 
> Given we don't have a versioning scheme, why fill in these optional
> fields?
> 
>> +RELEASE_TYPE="development"
> 
> Surely stable (the default) would be more accurate, given the examples
> for that include OpenSUSE Tumbleweed and Arch Linux, which have a
> similar rolling release model to what Spectrum will have?  We don't have
> a distinction between development and release builds, and I don't expect
> us to.

It's development because Spectrum OS is not yet stable.
Once Spectrum OS goes live it should obviously be changed.

>> +HOME_URL="https://www.spectrum-os.org/"
>> +VENDOR_URL="https://www.spectrum-os.org/"
> 
> "The VENDOR_NAME= field should be set if this one is"

Fair.

>> +ANSI_COLOR="1;34"
>> +DEFAULT_HOSTNAME="spectrum-host"
> 
> What do we expect this to do?

Causes systemd (if used) to set the hostname of the host.
-- 
Sincerely,
Demi Marie Obenour (she/her/hers)

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 7253 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]