setPointsToEdge should assert that its target isn't already merged,
but now that we batch up multiple merge requests, it's fine to allow
the target to be scheduled-for-merge.
Many assertions have been recently added and tightened in order to
"discover" unexpected cases. There's nothing incorrect about how these
cases were handled, but they lack unit tests. In this case I still
haven't been able to reduce a test case. I'm continuing to work on
it, but don't want to further delay the fix.
... including all SIL functions with are transitively referenced from such witness tables.
After the last devirtualizer run witness tables are not needed in the optimizer anymore.
We can delete witness tables with an available-externally linkage. IRGen does not emit such witness tables anyway.
This can save a little bit of compile time, because it reduces the amount of SIL at the end of the optimizer pipeline.
It also reduces the size of the SIL output after the optimizer, which makes debugging the SIL output easier.
LLVM ships a hardened memory allocator called Scudo:
https://llvm.org/docs/ScudoHardenedAllocator.html. This allocator
provides additional mitigations against heap-based vulnerabilities, but
retains sufficient performance to be safely run in production
applications.
While ideal Swift applications are obviously written in pure Swift, in
practice most applications contain some amount of code written in
less-safe languages. Additionally, plenty of Swift programs themselves
contain unsafe code, particularly when attempting to implement
high-performance data structures. These sources of unsafety introduce
the risk of memory issues, and having the option to use the Scudo
allocator is a useful defense-in-depth tool.
This patch enables `-sanitize=scudo` as an extra `swiftc` flag. This
sanitizer is only supported on Linux, so no further work is required to
enable it on Windows or Apple platforms. As this "sanitizer" is only a
runtime component, we do not require any wider changes to instrument
code. This is similar to clang's `-fsanitize=scudo` flag.
The Swift driver rejects platforms that don't support Scudo using an
existing mechanism in the Driver that is not part of this patch. This
mechanism is in swift::parseSanitizerArgValues(...)
(lib/Option/SanitizerOptions.cpp). The mechanism determines if a
sanitizer is supported by checking for the existence of the
corresponding sanitizer runtime library in the compiler's resource
directory. The Scudo runtime library currently only exists in the
Linux compiler resource directory. This results in the driver only
allowing Scudo when targeting Linux.
Effectively revert #28907. The request evaluator will also catch re-entrancy here, and those cycles can be broken with NameLookupFlags::IgnoreNewExtensions.
Soft revert a09382c. It should now be safe to add this flag back as an optimization to specifically disable lazy member loading instead of all extension loading.
Push the flag back everywhere it was needed, but also push it into lookup for associated type members which will never appear in extensions.
evaluable calls. The fix makes the check more conservative and
assumes that calls with generic arguments are not dead, as generic
functions with arbitrary side-effects can be invoked through them.
Also, add a few helper functions to the InstructionDeleter utility
that will enable deleting an instruction along with its users.
Add `SILFunctionType::getAutoDiffDerivativeFunctionType`.
It computes the derivative `SILFunctionType` for an "original"
`SILFunctionType`, given:
- Differentiability parameter indices
- Differentiability result index
- Derivative function kind
- Derivative function generic signature (optional)
- Other auxiliary parameters
Add doc comments explaining typing rules, preconditions, and other details.
Partially resolves TF-1124.
Unblocks upstreaming other SIL differentiable programming infrastructure.
