I had set up the driver to invoke a separate frontend invocation with
the "update code" mode. We sort of did this last release, except we
forked to the swift-update binary instead. This is causing problems with
testing in Xcode.
Instead, let's perform a single compile and add the remap file as an
additional output during normal compiles. The driver, seeing
-update-code, will add -emit-remap-file-path $PATH to the -c frontend
invocation.
rdar://problem/31857580
The Swift 4 Migrator is invoked through either the driver and frontend
with the -update-code flag.
The basic pipeline in the frontend is:
- Perform some list of syntactic fixes (there are currently none).
- Perform N rounds of sema fix-its on the primary input file, currently
set to 7 based on prior migrator seasons. Right now, this is just set
to take any fix-it suggested by the compiler.
- Emit a replacement map file, a JSON file describing replacements to a
file that Xcode knows how to understand.
Currently, the Migrator maintains a history of migration states along
the way for debugging purposes.
- Add -emit-remap frontend option
This will indicate the EmitRemap frontend action.
- Don't fork to a separte swift-update binary.
This is going to be a mode of the compiler, invoked by the same flags.
- Add -disable-migrator-fixits option
Useful for debugging, this skips the phase in the Migrator that
automatically applies fix-its suggested by the compiler.
- Add -emit-migrated-file-path option
This is used for testing/debugging scenarios. This takes the final
migration state's output text and writes it to the file specified
by this option.
- Add -dump-migration-states-dir
This dumps all of the migration states encountered during a migration
run for a file to the given directory. For example, the compiler
fix-it migration pass dumps the input file, the output file, and the
remap file between the two.
State output has the following naming convention:
${Index}-${MigrationPassName}-${What}.${extension}, such as:
1-FixitMigrationState-Input.swift
rdar://problem/30926261
- Add CompilerInvocation::getPCHHash
This will be used when creating a unique filename for a persistent
precompiled bridging header.
- Automatically generate and use a precompiled briding header
When we're given both -import-objc-header and -pch-output-dir
arguments, we will try to:
- Validate what we think the PCH filename should be for the bridging
header, based on the Swift PCH hash and the clang module hash.
- If we're successful, we'll just use it.
- If it's out of date or something else is wrong, we'll try to
emit it.
- This gives us a single filename which we can `stat` to check for the
validity of our code completion cache, which is keyed off of module
name, module filename, and module file age.
- Cache code completion results from imported modules
If we just have a single .PCH file imported, we can use that file as
part of the key used to cache declarations in a module. Because
multiple files can contribute to the __ObjC module, we've always given
it the phony filename "<imports>", which never exists, so `stat`-ing it
always fails and we never cache declarations in it.
This is extremely problematic for projects with huge bridging headers.
In the case where we have a single PCH import, this can bring warm code
completion times down to about 500ms from over 2-3s, so it can provide a
nice performance win for IDEs.
- Add a new test that performs two code-completion requests with a bridging header.
- Add some -pch-output-dir flags to existing SourceKit tests that import a bridging
header.
rdar://problem/31198982
This is purely designed to cheaply compute dependency graphs between
modules, and thus only lists the top-level names (i.e. not submodules)
and doesn't do any form of semantic analysis.
There's a class of errors in Serialization called "circularity
issues", where declaration A in file A.swift depends on declaration B
in file B.swift, and B also depends on A. In some cases we can manage
to type-check each of these files individually due to the laziness of
'validateDecl', but then fail to merge the "partial modules" generated
from A.swift and B.swift to form a single swiftmodule for the library
(because deserialization is a little less lazy for some things). A
common case of this is when at least one of the declarations is
nested, in which case a lookup to find that declaration needs to load
all the members of the parent type. This gets even worse when the
nested type is defined in an extension.
This commit sidesteps that issue specifically for nested types by
creating a top-level, per-file table of nested types in the "partial
modules". When a type is in the same module, we can then look it up
/without/ importing all other members of the parent type.
The long-term solution is to allow accessing any members of a type
without having to load them all, something we should support not just
for module-merging while building a single target but when reading
from imported modules as well. This should improve both compile time
and memory usage, though I'm not sure to what extent. (Unfortunately,
too many things still depend on the whole members list being loaded.)
Because this is a new code path, I put in a switch to turn it off:
frontend flag -disable-serialization-nested-type-lookup-table
https://bugs.swift.org/browse/SR-3707 (and possibly others)
Based off the PlaygroundTransform, this new ASTWalker leaves calls to __builtin_pc_before and __builtin_pc_after before and after a user would expect a program counter to enter a range of source code.
Given a source location, we can find the innermost enclosing scope
that describes that source location. Introduce this operation into the
scope map, then add a testing mode where we probe the scope map at
specifi locations to see what we find. Test for:
1) Finding the right innermost enclosing scope, and
2) That we're only expanding the part of the scope map that is needed
to identify that scope.
The scope map models all of the name lookup scopes within a source
file. It can be queried by source location to find the innermost scope
that contains that source location. Then, one can follow the parent
pointers in the scope to enumerate the enclosing scopes.
The scope map itself is lazily constructed, only creating scope map
nodes when required implicitly (e.g, when searching for a particular
innermost scope) or forced for debugging purposes.
using a lazily-constructed tree that can be searched by source
location. A search within a particular source location will
This is a /slightly/ more user-friendly option than
-debug-time-function-bodies; pass it a limit in milliseconds and
the compiler will warn whenever a function or multi-statement closure
takes longer than that to type-check.
Since it's a frontend option (and thus usually passed with -Xfrontend),
I went with the "joined" syntax as the common case. The usual "separate"
syntax of "-warn-long-function-bodies <N>" is also available.
As a frontend option, this is UNSUPPORTED and may be removed without
notice at any future date.
Additional caveats:
- Other parts of type-checking not measured by this may also be slow.
- May include first-use penalties (i.e. "this is slow because it's
the first function that references an imported type, which causes
many things to be imported")
- Does not report anything whatsoever about other phases of compilation
(SILGen, optimization, IRGen, assembly emission, whatever).
- Does not catch anything accidentally being type-checked multiple times
(a known issue for initial value expressions on properties).
We want to distinguish the special case of a library built with
-sil-serialize-all, from a SIL function that is [fragile] because
of an explicitly @_transparent or @inline(__always).
For now, NFC.
Since resilience is a property of the module being compiled,
not decls being accessed, we need to record which types are
resilient as part of the module.
Previously we would only ever look at the @_fixed_layout
attribute on a type. If the flag was not specified, Sema
would slap this attribute on every type that gets validated.
This is wasteful for non-resilient builds, because there
all types get the attribute. It was also apparently wrong,
and I don't fully understand when Sema decides to validate
which decls.
It is much cleaner conceptually to just serialize this flag
with the module, and check for its presence if the
attribute was not found on a type.
This times each phase of compilation, so you can see where time is being
spent. This doesn't cover all of compilation, but does get all the major
work being done.
Note that these times are non-overlapping, and should stay that way.
If we add more timers, they should go in a different timer group, so we
don't end up double-counting.
Based on a patch by @cwillmor---thanks, Chris!
Example output, from an -Onone build using a debug compiler:
===-------------------------------------------------------------------------===
Swift compilation
===-------------------------------------------------------------------------===
Total Execution Time: 8.7215 seconds (8.7779 wall clock)
---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Name ---
2.6670 ( 30.8%) 0.0180 ( 25.3%) 2.6850 ( 30.8%) 2.7064 ( 30.8%) Type checking / Semantic analysis
1.9381 ( 22.4%) 0.0034 ( 4.8%) 1.9415 ( 22.3%) 1.9422 ( 22.1%) AST verification
1.0746 ( 12.4%) 0.0089 ( 12.5%) 1.0834 ( 12.4%) 1.0837 ( 12.3%) SILGen
0.8468 ( 9.8%) 0.0171 ( 24.0%) 0.8638 ( 9.9%) 0.8885 ( 10.1%) IRGen
0.6595 ( 7.6%) 0.0142 ( 20.0%) 0.6737 ( 7.7%) 0.6739 ( 7.7%) LLVM output
0.6449 ( 7.5%) 0.0019 ( 2.6%) 0.6468 ( 7.4%) 0.6469 ( 7.4%) SIL verification (pre-optimization)
0.3505 ( 4.1%) 0.0023 ( 3.2%) 0.3528 ( 4.0%) 0.3530 ( 4.0%) SIL optimization
0.2632 ( 3.0%) 0.0005 ( 0.7%) 0.2637 ( 3.0%) 0.2639 ( 3.0%) SIL verification (post-optimization)
0.0718 ( 0.8%) 0.0021 ( 3.0%) 0.0739 ( 0.8%) 0.0804 ( 0.9%) Parsing
0.0618 ( 0.7%) 0.0010 ( 1.4%) 0.0628 ( 0.7%) 0.0628 ( 0.7%) LLVM optimization
0.0484 ( 0.6%) 0.0011 ( 1.5%) 0.0495 ( 0.6%) 0.0495 ( 0.6%) Serialization (swiftmodule)
0.0240 ( 0.3%) 0.0006 ( 0.9%) 0.0246 ( 0.3%) 0.0267 ( 0.3%) Serialization (swiftdoc)
0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) Name binding
8.6505 (100.0%) 0.0710 (100.0%) 8.7215 (100.0%) 8.7779 (100.0%) Total
Compute the hash of all interface tokens when parsing; write the
interface hash to the swiftdeps file, or if the -dump-interface-hash
option is passed to the frontend. This hash will be used in incremental
mode to determine whether a file's interface has changed, and therefore
whether dependent files need to be rebuilt in response to the change.
Committed on ChrisW's behalf while he gets his setup unborked.
rdar://problem/15352929
Swift SVN r30477
Instead of only honoring the last occurrence of -import-module, the frontend now
honors all occurrences of -import-module, making all of the modules specified on
the command line implicitly visible.
This fixes <rdar://problem/20422696>.
Swift SVN r27299
This flag indicates that internal APIs within the module should be made
available to client code for testing purposes. Currently does nothing.
Not ready for developer consumption yet, ergo a hidden frontend-only flag.
Part of testability (rdar://problem/17732115)
Swift SVN r26292
This will be needed for split-llvm code generation.
If multiple -o options are specified and only a single output file is needed
(currently always), the last one wins. This is NFC.
Swift SVN r25884
Separate InputFileKind from SourceFileKind, FrontendOptions will now use
InputFileKind, while Module will use SourceFileKind.
This is in preparation for adding an input file kind for LLVM IR.
rdar://19048891
Swift SVN r25555
This implicitly adds the named module as an import of every source file
in the module being compiled. This is not intended to be used generally,
but will be useful for playgrounds.
rdar://problem/19605934
Swift SVN r24905
There's also a testing option, -serialize-debugging-options, to force this
extra info to be serialized even for library targets. In the long run we'll
probably write out this information for all targets, but strip it out of
the "public module" when a framework is built. (That way it ends up in the
debug info's copy of the module.)
Incidentally, this commit includes the ability to add search paths to the
Clang importer on the fly, which is most of rdar://problem/16347147.
Unfortunately there's no centralized way to add search paths to both Clang
/and/ Swift at the moment.
Part of rdar://problem/17670778
Swift SVN r24545
This is a hidden frontend-only option intended for debugging purposes,
mainly for identifying where in a file the type checker is spending most
of its time. Use with "sort -g" to get the top problem functions.
Swift SVN r23789
This tracks top-level qualified and unqualified lookups in the primary
source file, meaning we see all top-level names used in the file. This
is part of the intra-module dependency tracking work that can enable
incremental rebuilds.
This doesn't quite cover all of a file's dependencies. In particular, it
misses cases involving extensions defined in terms of typealiases, and
it doesn't yet track operator lookups. The whole scheme is also very
dependent on being used to track file-level dependencies; if C is a subclass
of B and B is a subclass of A, C doesn't appear to depend on A. It only
works because changing A will mark B as dirty.
Part of rdar://problem/15353101
Swift SVN r22925
llvm::Optional lives in "llvm/ADT/Optional.h". Like Clang, we can get
Optional in the 'swift' namespace by including "swift/Basic/LLVM.h".
We're now fully switched over to llvm::Optional!
Swift SVN r22477
We do this so that the swiftmodule file contains all info necessary to
reconstruct the AST for debugging purposes. If the swiftmodule file is copied
into a dSYM bundle, it can (in theory) be used to debug a built app months
later. The header is processed with -frewrite-includes so that it includes
any non-modular content; the user will not have to recreate their project
structure and header maps to reload the AST.
There is some extra complexity here: a target with a bridging header
(such as a unit test target) may depend on another target with a bridging
header (such as an app target). This is a rare case, but one we'd like to
still keep working. However, if both bridging headers import some common.h,
we have a problem, because -frewrite-includes will lose the once-ness
of #import. Therefore, we /also/ store the path, size, and mtime of a
bridging header in the swiftmodule, and prefer to use a regular parse from
the original file if it can be located and hasn't been changed.
<rdar://problem/17688408>
Swift SVN r20128
This performs very conservative dependency generation for each compile task
within a full compilation. Any source file, swiftmodule, or Objective-C
header file that is /touched/ gets added to the dependencies list, which
is written out on a per-input basis at the end of compilation.
This does /not/ handle dependencies for the aggregated swiftmodule, swiftdoc,
generated header, or linked binary. This is just the minimum needed to get
Xcode to recognize what needs to be rebuilt when a header or Swift source
file changes. We can revisit this later.
This finishes <rdar://problem/14899639> for now.
Swift SVN r18045
Parse-only is a hot path; keep the semantics for it separate from normal parsing, otherwise it is very
easy to introduce something expensive without checking for Invocation.getParseOnly().
Also cleans up a bit CompilerInstance::performParse() as well.
Swift SVN r17596
