The bundling of the form of a request (e.g., the storage that makes up a request)
with the function that evaluates the request value requires us to perform
ad hoc indirection to address the AST —> Sema layering violation. For
example, ClassDecl::getSuperclass() calls through the LazyResolver (when
available) to form the appropriate request. This means that we cannot
use the the request-evaluator’s cache when LazyResolver is null, forcing
all cached state into the AST.
Provide the evaluator with a zone-based registration system, where each
request “zone” (e.g., the type checker’s requests) registers
callbacks to evaluate each kind of request within that zone. The
evaluator indirects through this table of function pointers, allowing
the request classes themselves to be available at a lower level (AST)
than the functions that perform the computation when the value isn’t
in the cache (e.g., Sema).
We are not taking advantage of the indirection yet; that’ll come in a
follow-up commit.
With more options coming for incremental syntax parsing, the list of
arguments will grow way to large and unhandy, so just extract them into
one common struct.
On the master-next branch, the InitLLVM class used by the PROGRAM_START
macro can modify the argc/argv values (currently only on Windows).
Expanding response files before initializing the stack trace also modifies
the arguments so use a separate SmallVector for each copy of the argument
vector.
This is a separate tool so we can use all of LLVM without worrying about
creating a dependency from the demangler on LLVM's libsupport.
I am going to use it to fix cmpcodesize for the new mangling by eliminating
cmpcodesize's usage of regex to classify symbols. Instead, we can just read in
the yaml version of the symbol trees for each demangled node and process that
instead.
rdar://41146023
Disable the sanitizers and code coverage when building a swift
invocation for the purpose of collecting diagnostics.
This should speed up diagnostic generation and reduce exposure to
compiler bugs.
rdar://40955900
When adding to the AST consumer queue, keep track of the snapshots
expected and only run such AST consumers when a new enough AST is built.
Progress is ensured because we always run the AST consumer that
triggered the build.
This prevents cases where enqueuing a consumer during an AST build has
different behaviour than enqueuing it after the AST has finished.
rdar://40340631
Introduce some metaprogramming of accessors and generally prepare
for storing less-structured accessor lists.
NFC except for a change to the serialization format.
Validation of the input side of FunctionTypeRepr was previously being done in Sema because of expression folding. If we instead push the invariant that the input TypeRepr should always be a TupleTypeRepr into the AST a number of nice cleanups fall out:
- The SIL Parser no longer accepts Swift 2-style type declarations
- Parse is more cleanly able to reject invalid FunctionTypeReprs
- Clients of the AST can be assured the input type is always a TupleType so we can flush Swift 2 hacks
This is a test-only change except for the introduction of a new request
that is only used by tests.
Our notifications are dispatched in the XPC event handler, which is not
synchronized with replies to explicit XPC send_message_with_reply calls.
This is fine for most users of sourcekitd, since the notifications are
already enqueued on the client side, but for testing we need a way to
guarantee that all notifications are passed to the client-side handler
before we exit. This commit introduces a new request for testing that
triggers a notification, allowing the client to wait on that
notification to ensure all previously posted notifications have been
handled.
Note: the non-deterministic test failures can be triggered by adding a
sleep of ~100 ms in the event handler before the notification is
dispatched to the main queue.
rdar://40311995
LLVM r334399 (and related Clang changes) moved clang::VersionTuple to
llvm::VersionTuple. Update Swift to match.
Patch by Jason Molenda.
rdar://problem/41025046
My previous change for this issue (033b884de6) did not fix all the
affected code. This gets the rest of them. Thanks for Jason Molenda
for helping with this. rdar://problem/41025365
instead of using name and decl context.
The advantages of this approach are three-fold:
- This is necessary to support inlined generic functions.
- We can retire the debugger-specific type name manfgling mode for archetypes.
- This saves 270kb of debug information in the x86_64 libSwiftCore.dylib alone.
<rdar://problem/38306256>
