The current implementation of using GetFinalPathNameByHandle actually reads all
intermediate links instead of just looking at the current link. This commit
alters the behavior of the function to use a different API which correctly reads
only one level of the soft link.
[breaking-change]
This PR makes `rustc` emit field names for tuple fields in DWARF. Formerly there was no way of directly accessing the fields of a tuple in GDB and LLDB since there is no C/C++ equivalent to this. Now, the debugger sees the name `__{field-index}` for tuple fields. So you can type for example `some_tuple_val.__2` to get the third tuple component.
When pretty printers are used (e.g. via `rust-gdb` or `rust-lldb`) these artificial field names will not clutter tuple rendering (which was the main motivation for not doing this in the past).
Solves #21948.
On linux, certain crashes can cause rust's backtrace generation code to consume huge amounts of memory (multiple gigabytes) when trying to generate a backtrace (e.g. servo/servo#5340).
I tracked this down to a bug in libbacktrace's allocation strategy when processing large binaries, and thankfully it's already been fixed upstream.
The change which fixes this specific issue is r210256, and is discussed here:
https://gcc.gnu.org/ml/gcc-patches/2014-05/msg00547.html
I assume we'd prefer to update libbacktrace wholesale, rather than cherry-picking individual patches, so this PR updates libbacktrace to the latest code from svn://gcc.gnu.org/svn/gcc/trunk/libbacktrace (r220112).
I've copied the upstream code and reapplied the parts of the following commits which touch src/libbacktrace:
- cd8f31759f
- fcb30a0b67
I have not reapplied f5765793b6 or d51047ded0, as those are both cosmetic cleanups applied to the entire codebase. For this directory, it seems worthwhile diverging as little as possible from upstream, to reduce the effort of future updates. But I can add them as well if preferred.
There are syntax extensions that call `std::rt::begin_unwind` passing it a `usize`. I updated the syntax extension to instead pass `u32`, but for bootstrapping reasons, I needed to create a `#[cfg(stage0)]` version of `std::rt::begin_unwind` and therefore also `panic!`.
Now that the internals of `format_args!` are unstable, tests that use it
don't compile after pretty-printing (unless they also declare the necessary
feature).
Arguments, Formatters, and the various format traits remain stable. The
format_args! macro uses #[allow_internal_unstable] to allow it access to
the unstable things in core::fmt.
Destabilized things include a "v1" in their name:
* core::fmt::rt
* core::fmt::rt::v1 (the module and all contents)
* core::fmt::ArgumentV1
* core::fmt::ArgumentV1::new
* core::fmt::ArgumentV1::from_usize
* core::fmt::Arguments::new_v1
* core::fmt::Arguments::new_v1_formatted
The unstable message was copied from that of std::io::_print.
`format_args!` uses `#[allow_internal_unstable]` to access internal
functions and structs that are marked unstable. For this to work, the
spans on AST nodes referencing unstable internals must be equal (same
lo/hi values) to the `format_args!` call site, so that the stability
checker can recognize that the AST node was generated by the macro.
Also add docs because not being able to discover these things is the pits.
This was lifted from http://blog.jgc.org/2015/04/the-one-line-you-should-add-to-every.html , and solves a problem I've solved myself a *lot* of times, often in massively inferior ways.
Super open to documenting it somewhere other than nitty-gritty, which isn't immediately obvious.
Right now, if the user requests to increase the vector size via reserve() or push_back() and the request brings the attempted memory above usize::MAX, we panic.
With this change there is only a panic if the minimum requested memory that could meet the requirement is above usize::MAX- otherwise it simply requests its largest capacity possible, usize::MAX.
It looks like `from_vec` was subsumed by new at some point,
but the documentation still refers to it as `from_vec`.
This updates the documentation for `from_vec_unchecked`
so that it properly says that it's the unchecked version of `new`.
Also, from_vec_unchecked requires a actual Vec<u8> while
new can take anything that is Into<Vec<u8>>, so I also
mention that in the documentation.
The "static and dynamic dispatch" chapter seems to no longer exist but there is a dead link from the Traits chapter pointing to it.
Have changed the link to point to "Trait Objects" which covers static and dynamic dispatch.
* In `noop_fold_expr`, call `new_span` in these cases:
- `ExprMethodCall`'s identifier
- `ExprField`'s identifier
- `ExprTupField`'s integer
Calling `new_span` for `ExprMethodCall`'s identifier is necessary to print
an acceptable diagnostic for `write!(&2, "")`. We see this error:
```
<std macros>:2:20: 2:66 error: type `&mut _` does not implement any method in scope named `write_fmt`
<std macros>:2 ( & mut * $ dst ) . write_fmt ( format_args ! ( $ ( $ arg ) * ) ) )
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
With this change, we also see a macro expansion backtrace leading to
the `write!(&2, "")` call site.
* After fully expanding a macro, we replace the expansion expression's
span with the original span. Call `fld.new_span` to add a backtrace to
this span. (Note that I'm call `new_span` after `bt.pop()`, so the macro
just expanded isn't on the backtrace.)
The motivating example for this change is `println!("{}")`. The format
string literal is `concat!($fmt, "arg")` and is inside the libstd macro.
We need to see the backtrace to find the `println!` call site.
* Add a backtrace to the `format_args!` format expression span.
r? alexcrichton
Addresses #23459
We provide tools to tell what exact symbols to emit for any fn or static, but
don’t quite check if that won’t cause any issues later on. Some of the issues
include LLVM mangling our names again and our names pointing to wrong locations,
us generating dumb foreign call wrappers, linker errors, extern functions
resolving to different symbols altogether (`extern {fn fail();} fail();` in some
cases calling `fail1()`), etc.
Before the commit we had a function called `note_unique_llvm_symbol`, so it is
clear somebody was aware of the issue at some point, but the function was barely
used, mostly in irrelevant locations.
Along with working on it I took liberty to start refactoring trans/base into
a few smaller modules. The refactoring is incomplete and I hope I will find some
motivation to carry on with it.
This is possibly a [breaking-change] because it makes dumbly written code
properly invalid.
This fixes all those issues about incorrect use of #[no_mangle] being not reported/misreported/ICEd by the compiler.
NB. This PR does not attempt to tackle the parallel codegen issue that was mentioned in https://github.com/rust-lang/rust/pull/22811, but I believe it should be very straightforward in a follow up PR by modifying `trans::declare::get_defined_value` to look at all the contexts.
cc @alexcrichton @huonw @nrc because you commented on the original RFC issue.
EDIT: wow, this became much bigger than I initially intended.
* In noop_fold_expr, call new_span in these cases:
- ExprMethodCall's identifier
- ExprField's identifier
- ExprTupField's integer
Calling new_span for ExprMethodCall's identifier is necessary to print
an acceptable diagnostic for write!(&2, ""). We see this error:
<std macros>:2:20: 2:66 error: type `&mut _` does not implement any method in scope named `write_fmt`
<std macros>:2 ( & mut * $ dst ) . write_fmt ( format_args ! ( $ ( $ arg ) * ) ) )
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
With this change, we also see a macro expansion backtrace leading to
the write!(&2, "") call site.
* After fully expanding a macro, we replace the expansion expression's
span with the original span. Call fld.new_span to add a backtrace to
this span. (Note that I'm call new_span after bt.pop(), so the macro
just expanded isn't on the backtrace.)
The motivating example for this change is println!("{}"). The format
string literal is concat!($fmt, "arg") and is inside the libstd macro.
We need to see the backtrace to find the println! call site.
* Add a backtrace to the format_args! format expression span.
Addresses #23459
This fixes the bug described in issue #23150. This affected formatting any floating point number into a string in a formatting pattern that: a) required rounding up, and b) required an extra digit on the front.
So `format!("{:.0}", 9.9)` would fail, but `format!("{:.0}", 8.9)` would succeed. This was due to a negative integer being cast to a `usize` resulting in an 'arithmetic operation overflowed' panic.
The fix was to change the order of operations so that the number is zero before casting.