Fix panic_fmt in the Book
While implementing panic_fmt for the GameCube I noticed that the parameters given to it were completely broken. Turns out that panic_fmt requires the C ABI to work correctly. This should be fixed in the documentation, so that others don't make the same mistake. Thanks to mbrubeck in the IRC for helping me figure this out.
Not specifying extern for lang_items correctly should potentially also be a compiler error.
This commit should help configure our OSX rustbuild builder for targeting 10.7.
A key part of this is using `libc++` instead of `libstdc++` as apparently it's
more filled out and otherwise LLVM's cmake configuration would fail.
mk: A few build fixes for i586-pc-windows-msvc
Detect the triple in the configure script for probing MSVC shenanigans and also
be sure to use `llvm-config` from the build host and not the target when
configuring compiler-rt.
mk: Fix cross-host builds
The change in b20e748 had the unintended consequence of breaking cross-host
builds as we apparently relied on the incorrect definition of this variable in
the makefiles. That change, however, was required to get tests passing so we
couldn't just revert it.
This commit fixes the underlying bug by leaving the "more correct" definition of
`LD_LIBRARY_PATH_ENV_TARGETDIR` (also fixing it with a hardcoded reference to
`CFG_BUILD`) and updating the `RPATH_VAR` definition below. Turned out we
already had special-casing logic for passing `--cfg stage1` during the
well-we-print-this-as-stage0 build of a cross-host. That logic was just updated
to pull from a different variable as opposed to relying on the definition of
that variable to accommodate this.
Closes#32568
Some blocks won't be translated at all because they aren't reachable at
the LLVM level, these need to be dealt with because they lack a
terminator and therefore trigger an LLVM assertion.
Other blocks aren't reachable because of codegen-time optimistions, for
example not dropping types that don't need it, often resulting in blocks
with no predecessors. We'll clean those up as well.
This is a fairly standard transform that inserts blocks along critical
edges so code can be inserted along the edge without it affecting other
edges. The main difference is that it considers a Drop or Call
terminator that would require an `invoke` instruction in LLVM a critical
edge. This is because we can't actually insert code after an invoke, so
it ends up looking similar to a critical edge anyway.
The transform is run just before translation right now.
Detect the triple in the configure script for probing MSVC shenanigans and also
be sure to use `llvm-config` from the build host and not the target when
configuring compiler-rt.
While implementing panic_fmt for the GameCube I noticed that the parameters given to it were completely broken. Turns out that panic_fmt requires the C ABI to work correctly. This should be fixed in the documentation, so that others don't make the same mistake. Thanks to mbrubeck in the IRC for helping me figure this out.
Not specifying extern for lang_items correctly should potentially also be a compiler error.
Use weak_odr linkage when reusing definitions across codegen units
When reuing a definition across codegen units, we obviously cannot use
internal linkage, but using external linkage means that we can end up
with multiple conflicting definitions of a single symbol across
multiple crates. Since the definitions should all be equal
semantically, we can use weak_odr linkage to resolve the situation.
Fixes#32518
r? @nikomatsakis
The change in b20e748 had the unintended consequence of breaking cross-host
builds as we apparently relied on the incorrect definition of this variable in
the makefiles. That change, however, was required to get tests passing so we
couldn't just revert it.
This commit fixes the underlying bug by leaving the "more correct" definition of
`LD_LIBRARY_PATH_ENV_TARGETDIR` (also fixing it with a hardcoded reference to
`CFG_BUILD`) and updating the `RPATH_VAR` definition below. Turned out we
already had special-casing logic for passing `--cfg stage1` during the
well-we-print-this-as-stage0 build of a cross-host. That logic was just updated
to pull from a different variable as opposed to relying on the definition of
that variable to accommodate this.
Closes#32568
Refactor s.t. TypeRelation implementors in `infer` don't escape InferCtxt
Some clean-up so that we can go back to the future of #31867 as opposed to #32542.
r? @nikomatsakis
When reuing a definition across codegen units, we obviously cannot use
internal linkage, but using external linkage means that we can end up
with multiple conflicting definitions of a single symbol across
multiple crates. Since the definitions should all be equal
semantically, we can use weak_odr linkage to resolve the situation.
Fixes#32518
style: Use `iter` for IntoIterator parameter names
This commit standardizes the codebase on `iter` for parameters with
IntoIterator bounds.
Previously about 40% of IntoIterator parameters were named `iterable`,
with most of the rest being named `iter`. There was a single place where
it was named `iterator`.
Prevent bumping the parser past the EOF.
Makes `Parser::bump` after EOF into an ICE, forcing callers to avoid repeated EOF bumps.
This ICE is intended to break infinite loops where EOF wasn't stopping the loop.
For example, the handling of EOF in `parse_trait_items`' recovery loop fixes#32446.
But even without this specific fix, the ICE is triggered, which helps diagnosis and UX.
This is a `[breaking-change]` for plugins authors who eagerly eat multiple EOFs.
See https://github.com/docopt/docopt.rs/pull/171 for such an example and the necessary fix.
melt the ICE when lowering an impossible range
Emit a fatal error instead of panicking when HIR lowering encounters a range with no `end` point.
This involved adding a method to wire up `LoweringContext::span_fatal`.
Fixes#32245 (cc @nodakai).
r? @nrc
