Remove RawVec::reserve_in_place.

[nnethercote]

And some related clean-ups.

r? @oli-obk

[nnethercote]

I recommend reviewing the commits one at a time.

This shouldn't affect perf, but I've seen small changes to RawVec::grow have surprisingly large effects, so let's do a perf run to check:
@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion

[bors]

⌛ Trying commit c6ac55dcbb3bc9a462147f4bd3530987b93c23fa with merge ce50223feb6be16fc4e3597602fb0326a1e89e3a...

[nnethercote]

@bors try

[bors]

⌛ Trying commit 6d514d1fe16b04ec3264c7bd3e396dda4d3c9bdd with merge 8d66106238885e2ad4a8da3600a2dd845483f66f...

[bors]

☀️ Try build successful - checks-azure
Build commit: 8d66106238885e2ad4a8da3600a2dd845483f66f (8d66106238885e2ad4a8da3600a2dd845483f66f)

[rust-timer]

Queued 8d66106238885e2ad4a8da3600a2dd845483f66f with parent 963bf52, future comparison URL.

[rust-timer]

Finished benchmarking try commit 8d66106238885e2ad4a8da3600a2dd845483f66f, comparison URL.

[nnethercote]

Perf results show a very slight win.

[nnethercote]

@Amanieu: with RawVec::reserve_in_place gone, there are no useful uses of ReallocPlacement::InPlace remaining in the codebase. It's only used in AllocRef::grow and AllocRef::shrink, and all implementations of those functions (the default impls in AllocRef, plus the impls in System and Global) simply return an error if InPlace is specified, even for shrink.

Perhaps ReallocPlacement should just be removed? In general InPlace only seems useful if you don't have a way of finding out the actual size of an allocation, in which case you might see if reallocing in place works. But AllocRef does provide a way to get the actual size of an allocation (in theory, if not in current practice), and being able to get the actual size is more useful in general.

[Amanieu]

reserve_in_place is used in src/libarena.

[Amanieu]

Oh nevermind, I see you just removes those uses.

[Amanieu]

cc @TimDiekmann

[TimDiekmann]

I don't like the idea to remove the feature to grow/shrink in-place from AllocRef. Actually I think this API is missing on Vec, too but wasn't implemented yet as GlobalAlloc does neither support growing in-place, nor overallocating. However ReallocPlacement is pretty long at typing and the order of the parameters in grow has to be looked up, which is why I don't like ReallocPlacement and it's siblings (anymore). There is a proposal around to revert it back do dedicated methods again: rust-lang/wg-allocators#58.

If the API is not going to be added to Vec it's possible to move in-place variants to a third-party crate, but this should be discussed in the linked issue.

[nnethercote]

8 days without review from @oli-obk, let's try someone else:

r? @Amanieu

[Amanieu]

Blocked on rust-lang/wg-allocators#58

[nnethercote]

Blocked on rust-lang/wg-allocators#58

It's not blocked... this PR removes a use of InPlace that was separate to the uses by AllocRef and its impls. It can land right now.

[Amanieu]

Oh right, good point.

@bors r+

[bors]

📌 Commit 6d514d1fe16b04ec3264c7bd3e396dda4d3c9bdd has been approved by Amanieu

[JohnTitor]

This will break rls and rustfmt builds so @bors r- while we're on the beta week.

   Compiling rustc-ap-arena v659.0.0
error[E0599]: no method named `reserve_in_place` found for struct `alloc::raw_vec::RawVec<T>` in the current scope
   --> /cargo/registry/src/git.luolix.top-1ecc6299db9ec823/rustc-ap-arena-659.0.0/lib.rs:228:39
    |
228 |                 if last_chunk.storage.reserve_in_place(currently_used_cap, n) {
    |                                       ^^^^^^^^^^^^^^^^ method not found in `alloc::raw_vec::RawVec<T>`

error[E0599]: no method named `reserve_in_place` found for struct `alloc::raw_vec::RawVec<u8>` in the current scope
   --> /cargo/registry/src/git.luolix.top-1ecc6299db9ec823/rustc-ap-arena-659.0.0/lib.rs:356:39
    |
356 |                 if last_chunk.storage.reserve_in_place(used_bytes, needed_bytes) {
    |                                       ^^^^^^^^^^^^^^^^ method not found in `alloc::raw_vec::RawVec<u8>`

error: aborting due to 2 previous errors

For more information about this error, try `rustc --explain E0599`.
[RUSTC-TIMING] rustc_ap_arena test:false 0.187
error: could not compile `rustc-ap-arena`.

To learn more, run the command again with --verbose.
warning: build failed, waiting for other jobs to finish...
[RUSTC-TIMING] jsonrpc_server_utils test:false 7.894
error: build failed
command did not execute successfully: "/checkout/obj/build/x86_64-unknown-linux-gnu/stage0/bin/cargo" "build" "--target" "x86_64-unknown-linux-gnu" "-Zbinary-dep-depinfo" "-j" "2" "--release" "--locked" "--color" "always" "--manifest-path" "/checkout/src/tools/rls/Cargo.toml" "--features" "clippy, rustc-workspace-hack/all-static" "--message-format" "json-render-diagnostics"
expected success, got: exit code: 101
[TIMING] ToolBuild { compiler: Compiler { stage: 2, host: "x86_64-unknown-linux-gnu" }, target: "x86_64-unknown-linux-gnu", tool: "rls", path: "src/tools/rls", mode: ToolRustc, is_optional_tool: true, source_type: Submodule, extra_features: ["clippy"] } -- 565.495
failed to test rls: could not build

...

error: Tool `rls` has regressed from test-pass to build-fail during beta week.
error: Tool `rustfmt` has regressed from test-pass to build-fail during beta week.

[bors]

☔ The latest upstream changes (presumably #72927) made this pull request unmergeable. Please resolve the merge conflicts.

[nnethercote]

I have rebased and beta week is over.

@bors r=Amanieu

[bors]

📌 Commit c9cbe7e has been approved by Amanieu

[bors]

⌛ Testing commit c9cbe7e with merge f8ed3f6c733f2119bda67b3cb01575c03005453b...

[Dylan-DPC-zz]

@bors retry (this pr is included in the rollup)