Following Clang's lead, and anecdotal evidence from the `float_one` part
of `run-make/extern-fn-struct-passing-abi`, use a floating point
register to return single-float aggregates, except on MSVC targets.
These fixes all have to do with the 64-bit PowerPC ELF ABI for big-endian
targets. The ELF v2 ABI for powerpc64le already worked well.
- Return after marking return aggregates indirect. Fixes#42757.
- Pass one-member float aggregates as direct argument values.
- Aggregate arguments less than 64-bit must be written in the least-
significant bits of the parameter space.
- Larger aggregates are instead padded at the tail.
(i.e. filling MSBs, padding the remaining LSBs.)
New tests were also added for the single-float aggregate, and a 3-byte
aggregate to check that it's filled into LSBs. Overall, at least these
formerly-failing tests now pass on powerpc64:
- run-make/extern-fn-struct-passing-abi
- run-make/extern-fn-with-packed-struct
- run-pass/extern-pass-TwoU16s.rs
- run-pass/extern-pass-TwoU8s.rs
- run-pass/struct-return.rs
Use hir::ItemLocalId instead of ast::NodeId in rustc::middle::region::CodeExtent.
This is an alternative to @michaelwoerister's #43887, changing `CodeExtent` instead of `ReScope`.
The benefit here is that the same `Region`s are used same-crate and cross-crate, while preserving the incremental recompilation properties of the stable `hir::ItemLocalId`.
Only places which needed to get back to the `ast::NodeId` from `CodeExtent` was its `span` method, used in error reporting - passing the `&RegionMaps` down allowed using `hir_to_node_id`.
`rustc::cfg` and `dataflow` also had to be converted to `hir::ItemLocalId` because of their interactions with `CodeExtent`, especially in `borrowck`, and from that we have 3 more `hir_to_node_id` calls: `cfg::graphviz` node labels, `borrowck` move reporting, and the `unconditional_recursion` lint.
Out of all of those, *only* the lint actually makes a decision (on whether code will compile) based on the result of the conversion, the others only use it to know how to print information to the user.
So I think we're safe to say that the bulk of the code working with a `CodeExtent` is fine with local IDs.
r? @nikomatsakis
add `fn` to syntax of rustc::ty::maps::define_maps
This is not a functional change, it just makes it possible to find a query by grepping without knowing that it's a query rather than a function.
I didn't pursue renaming everything from "map" to "query" because it seems to be a very invasive change. It would be a good test to exercise an IDE's renaming features.
Closes#44161
r? @eddyb
rustbuild: update the rust-src filter for compiler-rt
We wanted `src/compiler-rt/test` filtered from the `rust-src` package,
but that path is now `src/libcompiler_builtins/compiler-rt/test`. This
saves over half of the installed rust-src size. (50MB -> 22MB)
