Remove paste, simplify plugin macros #1069
Conversation
Bromeon
added
quality-of-life
No longer mangle the macro name. Since there are no more proc-macros (paste) involved, this should also speed up compile times in case of hundreds of class or #[itest] registrations.
Bromeon
force-pushed
the
qol/remove-paste
branch
from
180cc06 to
fb0d61c
Compare
|
API docs are being generated and will be shortly available at: https://godot-rust.github.io/docs/gdext/pr-1069 |
|
API docs are being generated and will be shortly available at: https://godot-rust.github.io/docs/gdext/pr-1069 |
| // As such, instead we export a fn with a random name of predictable format to be used | ||
| // by the embedder. | ||
| $crate::paste::paste! { | ||
| #[no_mangle] | ||
| extern "C" fn [< rust_gdext_registrant_ $gensym >] () { | ||
| __init(); | ||
| } |
There was a problem hiding this comment.
Was paste not still necessary here? We need a prefix with which to search for registered functions at init time.
There was a problem hiding this comment.
If needed, we can use pastey instead.
There was a problem hiding this comment.
Thanks, good catch! This is the problem when the comments aren't expressive: "a random name of predictable format" didn't make me think that the concrete prefix mattered.
If needed, we can use
pasteyinstead.
I don't simply trust random crates, and this one in particular had an incident where it just replaced all commits of the previous author. That after opening the RustSec advisory for paste, triggering the migration in the first place. Not a great start.
I'll see if I find another solution. Maybe a proc-macro, then we can get rid of gensym as well.
There was a problem hiding this comment.
Thanks for checking. Let's do it ourselves then.
Removes the
pastecrate, which is no longer maintained and started triggering cargo-deny alerts.The "plugin" macros are now simpler: they no longer mangle the registry name. Since there are no more proc-macros (
paste) involved, this should also speed up compile times in case of hundreds of class or#[itest]registrations.(I originally tried to do something very sophisticated with proc-macros, but abandoned ship as it grew too complex and I realized it not only wouldn't add value here, but actively make compilation slower. If ever we need those elsewhere, some efforts are preserved in e2aba76.)