On Linux it used to be possible to build only the stdlib, without the
SDK overlays, like so:
```
utils/build-script -- --build-swift-stdlib --build-swift-sdk-overlay=0
```
However this invocation now results in the following error:
```
+ /usr/bin/cmake --build /home/modocache/GitHub/apple/build/Ninja-ReleaseAssert/swift-linux-x86_64 -- -j8 all swift-stdlib-linux-x86_64
ninja: error: '/home/modocache/GitHub/apple/swift/stdlib/private/SwiftPrivatePthreadExtras/swiftGlibc-linux-x86_64', needed by 'stdlib/private/SwiftPrivatePthreadExtras/linux/x86_64/SwiftPrivatePthreadExtras.o', missing and no known rule to make it
utils/build-script: fatal error: command terminated with a non-zero exit status 1, aborting
```
The problem is that SwiftPrivatePthreadExtras is always built, regardless
of whether the SDK overlay is built. I believe there's an explicit check
against this for Darwin platforms to prevent the same error.
The solution, implemented here, is to add the same check for Linux.
* [stdlib] Adding new tests for collection indices
Changes:
- Added tests for 'index(next:)'
- Added tests for all 'formIndex' API variants
- Modified existing tests to better test bidirectional collections
* Fixes reflecting review feedback
As a first step to allowing the build script to build *only*
static library versions of the stdlib, change `add_swift_library`
such that callers must pass in `SHARED`, `STATIC`, or `OBJECT_LIBRARY`.
Ideally, only these flags would be used to determine whether to
build shared, static, or object libraries, but that is not currently
the case -- `add_swift_library` also checks whether the library
`IS_STDLIB` before performing certain additional actions. This will be
cleaned up in a future commit.
Due to a modeling error in the type checker's folding of type
references into type expressions, code such as "strideof(Int)" would
be accepted without the required ".self". Commit
4a60b6cbf4 fixes the modeling issue but
left the historical accepts-invalid; now, diagnose these cases with a
warning + Fix-It to ease the transition.
Fixes SR-899.
If a child process crashes outside of a test context, the parent process
should signal test failure. This behavior is a sign of something bad
happening in the child (for example, memory corruption), and should not
go unnoticed.
Child processes were exiting too early before the parent has a chance
to read a null pointer from the child, indicating that there are no
more instances to reflect. This wasn't a problem on OS X because the
I/O latency is so small compared to the iOS simulator, where the
problem would come up under heavy load. This makes the end-to-end
remote mirror tests deterministic again.
rdar://problem/26230879
... as well as new test collection types:
`MinimalRandomAccessCollectionWithStrideableIndex` and
`DefaultedRandonAccessCollectionWithStrideableIndex`, to test default
implementation of `index(...)` family of functions provided by the
standard library for the random access collections with strideable
indices.
Also add end-to-end tests for this finally, and fix a bug in
the SwiftReflectionTest library where we would give up on an
module completely if it did not have a field metadata section.
This is of course wrong if the module defines closures but
not nominal types.
