issue 1296: implement #[nolink]; deprecate #[link_name = ""] #1309
Conversation
…emove empty link_name. Can't remove them from stdlib until the snapshotted compiler supports #[nolink].
|
This looks great so far. Thanks! |
|
I have a question. Cstore::add_used_library() returns false if a library has already been added by another module. Currently, creader::visit_item() returns early if this is the case, skipping the link_args section. This will change if I rework it to address #1295 as well. I'm thinking that if two modules specify a library, it's risky to let the second one specify link_args -- they may duplicate or conflict with the first module's link_args. So it probably should throw an error if link_args are specified for a redundant library, or at least generate a warning. Thoughts? |
…same lib, and second has link_args. I think it should undefined to have multiple modules that link in the same library, but provide different link arguments. Unfortunately we don't track link_args by module -- they are just appended as discovered into the crate store -- but for now, it should be an error to provide link_args on a module that's already been included (with or without link_args).
|
Either a warning or error if two native modules with the same name specify different link_args sounds reasonable to me. |
|
Thanks, brson, this was fun. :) |
|
The Is |
These are the last remaining platform intrinsics necessary for portable-simd.
* Remove everything related to DocContext * Remove everything about rustdoc Options * remove everything related to rustdoc cache * clean up markdown * remove clean, fold, and visit from librustdoc * remove config, core, lint * Cull cargo deps from librustdoc
The #[link_name = ""] attributes in the standard library will need to be removed later, after new compiler snapshots are generated; I've left fixme notes to take care of that later.