first pass at isolating dep-graph tasks
This intentionally leaves `DepGraph::in_task()`, the more common form,
alone. Eventually all uses of `DepGraph::in_task()` should be ported
to `with_task()`, but I wanted to start with a smaller subset.
I also used `AssertDepGraphSafe` on the closures that are found in
trans. This is because the types there are non-trivial and I wanted to
lay down the mechanism and come back to the more subtle cases.
The current approach taken in this PR has a downside: it is necessary
to manually "reify" fn types into fn pointers when starting a task,
like so:
dep_graph.with_task(..., task_fn as fn(_))
this is because `with_task` takes some type `T` that implements
`DepGraphTask` rather than taking a `fn()` type directly. *This* is so
that we can accept closure and also so that we can accept fns with
multiple arities. I am not sure this is the right approach.
Originally I wanted to use closures bound by an auto trait, but that
approach has some limitations:
- the trait cannot have a `read()` method; since the current method
is unused, that may not be a problem.
- more importantly, we would want the auto trait to be "undefined" for all types
*by default* -- that is, this use case doesn't really fit the typical
auto trait scenario. For example, imagine that there is a `u32` loaded
out of a `hir::Node` -- we don't really want to be passing that
`u32` into the task!
This is a last-ditch attempt to help our pain with dealing with #38878 on the
bots. A new environment variable is added to the compiler,
`RUSTC_RETRY_LINKER_ON_SEGFAULT`, which will instruct the compiler to
automatically retry the final linker invocation if it looks like the linker
segfaulted (up to 2 extra times).
Unfortunately there have been no successful attempts to debug #38878. The only
information seems to be that the linker (e.g. `ld` on OSX) is segfaulting
somewhere in some thread pool implementation. This appears to be spurious as
failed PRs will later merge.
The hope is that this helps the queue keep moving without clogging and delaying
PRs due to #38878.
A task function is now given as a `fn` pointer to ensure that it carries
no state. Each fn can take two arguments, because that worked out to be
convenient -- these two arguments must be of some type that is
`DepGraphSafe`, a new trait that is intended to prevent "leaking"
information into the task that was derived from tracked state.
This intentionally leaves `DepGraph::in_task()`, the more common form,
alone. Eventually all uses of `DepGraph::in_task()` should be ported
to `with_task()`, but I wanted to start with a smaller subset.
Originally I wanted to use closures bound by an auto trait, but that
approach has some limitations:
- the trait cannot have a `read()` method; since the current method
is unused, that may not be a problem.
- more importantly, we would want the auto trait to be "undefined" for all types
*by default* -- that is, this use case doesn't really fit the typical
auto trait scenario. For example, imagine that there is a `u32` loaded
out of a `hir::Node` -- we don't really want to be passing that
`u32` into the task!
Let `-Crelocation-model` better control `-pie` linking
Prior to this, if relocation model in the target options was "pic", as
most targets have it, then the user's `-Crelocation-model` had no effect
on whether `-pie` would be used. Only `-Clink-arg=-static` would have a
further override here.
Now we use `context::get_reloc_model`, which gives precedence to the
user's option, and if it's `RelocMode::PIC` we enable `-pie`. This is
the same test as `context::is_pie_binary`, except that folds across all
`sess.crate_types`, not just the current one.
Fixes#35061.
Prior to this, if relocation model in the target options was "pic", as
most targets have it, then the user's `-Crelocation-model` had no effect
on whether `-pie` would be used. Only `-Clink-arg=-static` would have a
further override here.
Now we use `context::get_reloc_model`, which gives precedence to the
user's option, and if it's `RelocMode::PIC` we enable `-pie`. This is
the same test as `context::is_pie_binary`, except that folds across all
`sess.crate_types`, not just the current one.
convert AdtDef::destructor to on-demand
This removes the `Cell` from `AdtDef`. Also, moving destructor validity
checking to on-demand (forced during item-type checking) ensures that
invalid destructors can't cause ICEs.
Fixes#38868.
Fixes#40132.
r? @eddyb
Tracking issue: https://github.com/rust-lang/rust/issues/40180
This calling convention can be used for definining interrupt handlers on
32-bit and 64-bit x86 targets. The compiler then uses `iret` instead of
`ret` for returning and ensures that all registers are restored to their
original values.
Usage:
```
extern "x86-interrupt" fn handler(stack_frame: &ExceptionStackFrame) {…}
```
for interrupts and exceptions without error code and
```
extern "x86-interrupt" fn page_fault_handler(stack_frame: &ExceptionStackFrame,
error_code: u64) {…}
```
for exceptions that push an error code (e.g., page faults or general
protection faults). The programmer must ensure that the correct version
is used for each interrupt.
For more details see the [LLVM PR][1] and the corresponding [proposal][2].
[1]: https://reviews.llvm.org/D15567
[2]: http://lists.llvm.org/pipermail/cfe-dev/2015-September/045171.html
This removes the Cell from AdtDef. Also, moving destructor validity
checking to on-demand (forced during item-type checking) ensures that
invalid destructors can't cause ICEs.
Fixes#38868.
Fixes#40132.
The types of statics, like all other items, are stored in the tcx
unnormalized. This is necessarily so, because
a) Item types other than statics have generics, which can't be
normalized.
b) Eager normalization causes undesirable on-demand dependencies.
Keeping with the principle that MIR lvalues require no normalization in
order to interpret, this patch stores the normalized type of the statics
in the Lvalue and reads it to get the lvalue type.
Fixes#39367.
trans: don't ICE when trying to create ADT trans-items
ADTs are translated in-place from rustc_trans::callee, so no trans-items
are needed.
This fix will be superseded by the shimmir branch, but I prefer not to
backport that to beta.
Fixes#39823.
Beta-nominating because regression.
r? @michaelwoerister
add `-C overflow-checks` option
In addition to defining and handling the new option, we also add a method on librustc::Session for determining the necessity of overflow checks. This method provides a single point to sort out the three (!) different ways for turning on overflow checks: -C debug-assertions, -C overflow-checks, and -Z force-overflow-checks.
I was seeing a [run-pass/issue-28950.rs](b1363a73ed/src/test/run-pass/issue-28950.rs) failure on my machine with these patches, but I was also seeing the failure without the changes to the core compiler. We'll see what travis says.
Fixes#33134. r? @alexcrichton
Set metadata for vtable-related loads
Give LLVM much more information about vtable pointers. Without the extra
information, LLVM has to be rather pessimistic about vtables, preventing
a number of obvious optimisations.
* Makes the vtable pointer argument noalias and readonly.
* Marks loads of the vtable pointer as nonnull.
* Marks load from the vtable with `!invariant.load` metadata.
Fixes#39992
ADTs are translated in-place from rustc_trans::callee, so no trans-items
are needed.
This fix will be superseded by the shimmir branch, but I prefer not to
backport that to beta.
Fixes#39823.
In addition to defining and handling the new option, we also add a
method on librustc::Session for determining the necessity of overflow
checks. This method provides a single point to sort out the three (!)
different ways for turning on overflow checks: -C debug-assertions, -C
overflow-checks, and -Z force-overflow-checks.
Fixes#33134.
Give LLVM much more information about vtable pointers. Without the extra
information, LLVM has to be rather pessimistic about vtables, preventing
a number of obvious optimisations.
* Makes the vtable pointer argument noalias and readonly.
* Marks loads of the vtable pointer as nonnull.
* Marks load from the vtable with `!invariant.load` metadata.
Fixes#39992
When targeting Emscripten, rustc emits Vector Instructions by default.
However Web Assembly doesn't support Vector Instructions yet, which
causes Binaryen to fail converting the intermediate asm.js code to Web
Assembly. While asm.js kind of supports Vector Instructions, they
aren't supported by any browser other than Firefox, often meaning that
they need to be emulated very slowly. So it should just be turned off
for all Emscripten targets.
Fixes#38558
erase late bound regions in `get_vtable_methods()`
Higher-ranked object types can otherwise cause late-bound regions to
sneak into the substs, leading to the false conclusion that some method
is unreachable.
r? @arielb1, who wrote the heart of this patch anyhow
Fixes#39292
Higher-ranked object types can otherwise cause late-bound regions to
sneak into the substs, leading to the false conclusion that some method
is unreachable. The heart of this patch is from @arielb1.
[MIR] SwitchInt Everywhere
Something I've been meaning to do for a very long while. This PR essentially gets rid of 3 kinds of conditional branching and only keeps the most general one - `SwitchInt`. Primary benefits are such that dealing with MIR now does not involve dealing with 3 different ways to do conditional control flow. On the other hand, constructing a `SwitchInt` currently requires more code than what previously was necessary to build an equivalent `If` terminator. Something trivially "fixable" with some constructor methods somewhere (MIR needs stuff like that badly in general).
Some timings (tl;dr: slightly faster^1 (unexpected), but also uses slightly more memory at peak (expected)):
^1: Not sure if the speed benefits are because of LLVM liking the generated code better or the compiler itself getting compiled better. Either way, its a net benefit. The CORE and SYNTAX timings done for compilation without optimisation.
```
AFTER:
Building stage1 std artifacts (x86_64-unknown-linux-gnu -> x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 31.50 secs
Finished release [optimized] target(s) in 31.42 secs
Building stage1 compiler artifacts (x86_64-unknown-linux-gnu -> x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 439.56 secs
Finished release [optimized] target(s) in 435.15 secs
CORE: 99% (24.81 real, 0.13 kernel, 24.57 user); 358536k resident
CORE: 99% (24.56 real, 0.15 kernel, 24.36 user); 359168k resident
SYNTAX: 99% (49.98 real, 0.48 kernel, 49.42 user); 653416k resident
SYNTAX: 99% (50.07 real, 0.58 kernel, 49.43 user); 653604k resident
BEFORE:
Building stage1 std artifacts (x86_64-unknown-linux-gnu -> x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 31.84 secs
Building stage1 compiler artifacts (x86_64-unknown-linux-gnu -> x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 451.17 secs
CORE: 99% (24.66 real, 0.20 kernel, 24.38 user); 351096k resident
CORE: 99% (24.36 real, 0.17 kernel, 24.18 user); 352284k resident
SYNTAX: 99% (52.24 real, 0.56 kernel, 51.66 user); 645544k resident
SYNTAX: 99% (51.55 real, 0.48 kernel, 50.99 user); 646428k resident
```
cc @nikomatsakis @eddyb
This removes another special case of Switch by replacing it with the more general SwitchInt. While
this is more clunky currently, there’s no reason we can’t make it nice (and efficient) to use.
