Re: [PATCH 0/8] rust: io: register: allow paths for relative register bases
"Gary Guo" <[email protected]> Tue, 21 Jul 2026 16:06:04 +0100
| Newsgroups | dev.linux.lists.nova-gpu,dev.linux.lists.driver-core,org.freedesktop.lists.dri-devel,org.kernel.vger.linux-kernel,org.kernel.vger.rust-for-linux |
|---|---|
| Message-ID | <[email protected]> |
On Tue Jul 21, 2026 at 4:02 PM BST, Alexandre Courbot wrote: > On Tue Jul 21, 2026 at 7:34 AM PDT, Gary Guo wrote: >> On Tue Jul 21, 2026 at 12:00 PM BST, Alexandre Courbot wrote: >>> The register! macro currently accepts an identifier as the base of a >>> relative register. This requires base types declared in other modules to >>> be imported into the module containing the register declaration, rather >>> than allowing a qualified path to be used directly, which is arbitrarily >>> limiting. >>> >>> This series allows path to be used as relative register bases. Since a >>> `path` fragment cannot be followed by `+` in declarative macros, the >>> `Base + Offset` and `Base + Alias` forms are replaced with `Base: >>> Offset` and `Base: Alias`, inducing a syntax change. >>> >>> The first two patches clean up the macro's internal dispatch and >>> arguments in preparation for the change. Patches 3-6 gradually add >>> support for the new syntax and update docs and examples, while >>> preserving the old syntax. Patch 7 updates Nova to use the new syntax, >>> and patch 8 removes support for the now unused old syntax. >>> >>> The series is structured this way so the Nova update and deprecated >>> syntax removal can be merged separately, as otherwise the whole series >>> would need to go through drm-rust. >> >> I would like to avoid adding new features to relative registers as I aspire to >> get rid of them entirely with typed registers. > > This is a fix rather than a new feature. Relative register bases should > have been captured as a path since the beginning. > > I'll be glad to remove relative registers entirely if projections can > cover their use-case better, but I would not withhold a fix while a > replacement is not visible yet. I'm not seeing this as a fix. Nova doesn't need to use a path, and requiring an import is hardly a restriction. Best, Gary