On Linux/Mac, I got a segmentation fault:
(gdb) bt
#0 0x00000000007519af in glue_take584 ()
#1 0x00000000006d4bec in
back::rpath::get_rpath_flags::_3899df2ca513c603 ()
#2 0x00000000006c7655 in back:🔗:link_binary::_7afde00a9791031c ()
#3 0x00000000007d3ff5 in driver::rustc::compile_input::thunk9212 ()
#4 0x0000000000710f24 in driver::rustc::time::_3e691b2a4ba58aee ()
#5 0x000000000071a79d in
driver::rustc::compile_input::_7b4a41b87c18e034 ()
#6 0x000000000072f0a9 in driver::rustc::main::_cd8b8c8185af3dee ()
#7 0x000000000072f1ed in _rust_main ()
#8 0x00007ffff7e6e146 in task_start_wrapper (a=<optimized out>) at
../src/rt/rust_task.cpp:176
The variable `output` or `out_filename` becomes (null) after the definition
of `fn unlib`. Move the function defintion to the beginning seems
prevent the crash on Linux.
[crate_type = "lib"] builds it as a library.
[crate_type = "bin"] builds it as an executable.
Executable is the default. --lib and --bin switches override.
See src/test/run-pass/nested-patterns.rs for some examples. The syntax is
boundvar@subpattern
Which will match the subpattern as usual, but also bind boundvar to the
whole matched value.
Closes#838
Also shuffles around the organization of numeric literals and types,
separating by int/uint/float instead of machine-vs-non-machine types.
This simplifies some code.
Closes#974Closes#1252
When unwinding through __morestack the stack limit in the TLS is invalidated
and must be reset. Instead of actually landing at __morestack we're
just going to make all our Rust landing pads call upcall_reset_stack_limit,
which will find the stack segment that corresponds to the current stack
pointer and put the limit in the TLS.
Also massively expand the stack segment red zone to make more room for the
dynamic linker. Will fix in the future.
LLVM code generator emits the ".file filename" directive for ELF
backends. Value of the "filename" is set as the LLVM module identifier.
Due to a LLVM MC bug[1], LLVM crashes if the module identifer is same as
other symbols such as a function name in the module.
This patch adds a ".rc" suffix (means crates) to LLVM module identifier
to workaround the bug.
Fixes issue #1251.
1. http://llvm.org/bugs/show_bug.cgi?id=11479
The path information was an optional "filename" component of crate
directive AST. It is now replaced by an attribute with metadata named
"path".
With this commit, a directive
mod foo = "foo.rs";
should be written as:
#[path = "foo.rs"]
mod foo;
Closes issue #906.
Temporarily allow path specified in either as attribute or in AST, like:
#[path = "mymod.rs"]
mod mymod = "mymod.rs";
This is a transitional commit to avoid creating a stage1 snapshot.
Diagnostic highlight lines are incorrect placed when the related line
number is 10, 100, etc.
The root cause is line number are treated as 0 based (should be 1 based)
when calculating offset of line number digits.
It's proving too inflexible, so I'm ripping out the extra complexity
in the hope that regions will, at some point, provide something
similar.
Closes#918
The reference now has an empty hole where the auth keyword used to be.
Changing the keyword table seems to require manually sorting the
keywords and putting them back into some kind of arcane interleaved
order. I'll open an issue to actually fix this.
Closes#1211
Not included in the build by default, since it's fragile and kludgy. Do
something like this to run it:
cd doc/tutorial
RUSTC=../../build/stage2/bin/rustc bash test.sh
Closes#1143
This involved adding 'copy' to more generics than I hoped, but an
experiment with making it implicit showed that that way lies madness --
unless enforced, you will not remember to mark functions that don't
copy as not requiring copyable kind.
Issue #1177
Doing something like some([1, 2, 3]) will now no longer create a temporary
copy of the vector. It will also be easier for the kind checker to see that
putting a resource into a data-structure constructor is safe.
This is intended to solve the problem of how to pass arguments to
constructor functions -- you want to move in rvalues, but not have to
explicitly copy stuff that is not an rvalue. The by-copy passing
convention will ensure the callee gets its own copy of the value. For
rvalues, it'll just pass off the value. For lvalues, it'll make a
copy.
Issue #1177
This goes before a snapshot, so that subsequenct patches can make the
transition without breaking the build. Disables kind checking pass, makes
parser accept both new and old-style kind annotation.
Issue #1177
