On 1/28/25 15:38, Thomas Lamprecht wrote:
This is a bit worded like that behavior would be a regression, but it
isn't AFAICT as this was always kebab-case from when being added in
commit 6526662 ("fix #5579: auto-installer: add optional first-boot hook
script"); or am I overlooking something?
I'm sorry that the commit message came across like this, I didn't intend
to word it as a regression, but I can see why it did. I wasn't aware
that we prefer kebab-case for newer property names in the answer file
and I'll keep that in mind for future patches.
But we prefer kebab-case for any public API/CLI parameter for modern code;
so shouldn't we rather to the opposite, transform all other (de)serializable
configs to use kebab-case with backward-compat aliases for the cases it
matters?
I also like that solution and that is more in line with the motivation
behind the patch. I could queue up a patch for the next ISO release, so
that it's indifferent to the user whether they write the config
parameter names in their answer files in the old snake case or new
kebab-case format.
I'd prefer a single serde attribute for that rather than rename_all +
alias at every property, if that's possible in a few lines, as it would
be cleaner and we don't have to look at every property whether it's
missing a alias attribute or not.
_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
