The arguments were accidentally swapped in the wrong order.

Closes #12743

This leverages the new hashing framework and hashmap implementation to provide a
much speedier hashing algorithm for node ids and def ids. The hash algorithm
used is currentl FNV hashing, but it's quite easy to swap out.

I originally implemented hashing as the identity function, but this actually
ended up in slowing down rustc compiling libstd from 8s to 13s. I would suspect
that this is a result of a large number of collisions.

With FNV hashing, we get these timings (compiling with --no-trans, in seconds):

|           |  before  |  after  |
|-----------|---------:|--------:|
| libstd    |   8.324  |  6.703  |
| stdtest   |  47.674  | 46.857  |
| libsyntax |   9.918  |  8.400  |

If #[feature(default_type_parameters)] is enabled for a crate, then
deriving(Hash) will expand with Hash<W: Writer> instead of Hash<SipState> so
more hash algorithms can be used.

The arguments were accidentally swapped in the wrong order.

Closes #12743

This leverages the new hashing framework and hashmap implementation to provide a
much speedier hashing algorithm for node ids and def ids. The hash algorithm
used is currentl FNV hashing, but it's quite easy to swap out.

I originally implemented hashing as the identity function, but this actually
ended up in slowing down rustc compiling libstd from 8s to 13s. I would suspect
that this is a result of a large number of collisions.

With FNV hashing, we get these timings (compiling with --no-trans, in seconds):

|           |  before  |  after  |
|-----------|---------:|--------:|
| libstd    |   8.324  |  6.703  |
| stdtest   |  47.674  | 46.857  |
| libsyntax |   9.918  |  8.400  |

Closes #6738
Closes #7061
Closes #7899
Closes #9719
Closes #10028
Closes #10228
Closes #10401
Closes #11192
Closes #11508
Closes #11529
Closes #11873
Closes #11925

Turns out sundown has already escaped this content for us, so there's no need
for us to escape it again.

Closes #12736

On Windows, `LLVMRustGetLastError()` may return non-utf8 mojibake string
if system uses non-English locale. It caused ICE when llvm fails.
This patch doesn't fix the real problem, but just make rustc not die.

Closes #6738
Closes #7061
Closes #7899
Closes #9719
Closes #10028
Closes #10228
Closes #10401
Closes #11192
Closes #11508
Closes #11529
Closes #11873
Closes #11925

Turns out sundown has already escaped this content for us, so there's no need
for us to escape it again.

Closes #12736

On Windows, `LLVMRustGetLastError()` may return non-utf8 mojibake string
if system uses non-English locale. It caused ICE when llvm fails.
This patch doesn't fix the real problem, but just make rustc not die.

Cosmetic changes at best, but there are so many such typos that I couldn't ignore them. :) Some occurrences of typos are linked to the generated documentations but no changes should break the builds.

This version is slightly more up to date and is closer to the 3.5 that we're
using. This also updates the travis config to have a build matrix which tests
rust against LLVM 3.3 and 3.4. For pull requests only LLVM 3.4 is tested to
reduce the load on travis.

This is mostly just fluff, there's no real reason to gate rust on these results,
it's more of just a nice thing to know when we break compatibility with LLVM 3.3
and 3.4 (and eventually 3.5). This turns off notifications of failed commits
(which are sent out for pushes to master).

The llvm.copysign and llvm.round intrinsics weren't added until LLVM 3.4, so if
we're on LLVM 3.3 we lower these to calls in libm instead of LLVM intrinsics.

This should fix our travis failures.

This PR brings back limited debuginfo which allows for nice backtraces and breakpoints, but omits any info about variables and types.

The `-g` and `--debuginfo` command line options have been extended to take an optional argument:
`-g0` means no debug info.
`-g1` means line-tables only.
`-g2` means full debug info.

Specifying `-g` without argument is equivalent to `-g2`.

Fixes #12280

Details are in the commit messages, but this closes a few issues seen with `libnative` recently.

This version is slightly more up to date and is closer to the 3.5 that we're
using. This also updates the travis config to have a build matrix which tests
rust against LLVM 3.3 and 3.4. For pull requests only LLVM 3.4 is tested to
reduce the load on travis.

This is mostly just fluff, there's no real reason to gate rust on these results,
it's more of just a nice thing to know when we break compatibility with LLVM 3.3
and 3.4 (and eventually 3.5). This turns off notifications of failed commits
(which are sent out for pushes to master).

This commit adds a appear-on-over link to all section headers to generated
documentation. Each header also receives an id now, even those generated through
markdown. The purpose of this is to provide easy to link to sections.

This modifies the default header markdown generation because the default id
added looks like "toc_NN" which is difficult to reconcile among all sections (by
default each section gets a "toc_0" id), and it's also not very descriptive of
where you're going.

This chooses to adopt the github-style anchors by taking the contents of the
title and hyphen-separating them (after lower casing).

Closes #12681

This commit adds a appear-on-over link to all section headers to generated
documentation. Each header also receives an id now, even those generated through
markdown. The purpose of this is to provide easy to link to sections.

This modifies the default header markdown generation because the default id
added looks like "toc_NN" which is difficult to reconcile among all sections (by
default each section gets a "toc_0" id), and it's also not very descriptive of
where you're going.

This chooses to adopt the github-style anchors by taking the contents of the
title and hyphen-separating them (after lower casing).

Closes #12681

Closes #8506.

When using tasks in Rust, the expectation is that the runtime does not exit
before all tasks have exited. This is enforced in libgreen through the
`SchedPool` type, and it is enforced in libnative through a `bookkeeping` module
and a global count/mutex pair. Unfortunately, this means that a process which
originates with libgreen will not wait for spawned native tasks.

In order to fix this problem, the bookkeeping module was moved from libnative to
libstd so the runtime itself can wait for native tasks to exit. Green tasks do
not manage themselves through this bookkeeping module, but native tasks will
continue to manage themselves through this module.

Closes #12684

Closes #8506.

The `trans::adt` code for statics uses fields with `C_undef` values to
insert alignment padding (because LLVM's own alignment padding isn't
always sufficient for aggregate constants), and assumes that all fields
in the actual Rust value are represented by non-undef LLVM values, to
distinguish them from that padding.

But for nullable pointer enums, if non-null variant has fields other
than the pointer used as the discriminant, they would be set to undef in
the null case, to reflect that they're never accessed.

To avoid the obvious conflict between these two items, the latter undefs
were wrapped in unary LLVM structs to distinguish them from the former
undefs.  Except this doesn't actually work -- LLVM, not unreasonably,
treats the "wrapped undef" as a regular undef.

So this commit just sets all fields to null in the null pointer case of
a nullable pointer enum static, because the other fields don't really
need to be undef in the first place.

The llvm.copysign and llvm.round intrinsics weren't added until LLVM 3.4, so if
we're on LLVM 3.3 we lower these to calls in libm instead of LLVM intrinsics.

This should fix our travis failures.

The `cmp::min` and `cmp::max` functions are not correct with partially
ordered values. #12712

Fixes #10877

There was another PR which attempted to fix this in the parser (#11804) and which was closed due to inactivity.
This PR modifies typeck instead (as suggested in #11804), which indeed seems to be simpler than modifying the parser and allows for a better error message.

Closes #10877

- Moves mtwt hygiene code into its own file
- Fixes FIXME's which leads to ~2x speed gain in expansion pass
- It is now @-free

Just realized that these were quite outdated and quickly fixed it. I have to admit that I did not run the test suite on this one though…

When the timer_helper thread exited, it would attempt to re-acquire the global
task count mutex, but the mutex had previously been deallocated, leading to
undefined behavior of the mutex, and in some cases deadlock.

Another mutex is used to coordinate shutting down the timer helper thread.

Closes #12699

Using nanosleep() allows us to gracefully recover from EINTR because on error it
fills in the second parameter with the remaining time to sleep.

Closes #12689

This function is not threadsafe, and is deprecated in favor of the threadsafe
readdir_r variant.

Closes #12692