the cases where we can't determine whether special members would be trivial
while building the class, we eagerly declare those special members. The impact
of this is bounded, since it does not trigger implicit declarations of special
members in classes which merely *use* those classes.
In order to determine whether we need to apply this rule, we also need to
eagerly declare move operations and destructors in cases where they might be
deleted. If a move operation were supposed to be deleted, it would instead
be suppressed, and we could need overload resolution to determine if we fall
back to a trivial copy operation. If a destructor were implicitly deleted,
it would cause the move constructor of any derived classes to be suppressed.
As discussed on cxx-abi-dev, C++11's selected constructor rules are also
retroactively applied as a defect resolution in C++03 mode, in order to
identify that class B has a non-trivial copy constructor (since it calls
A's constructor template, not A's copy constructor):
struct A { template<typename T> A(T &); };
struct B { mutable A a; };
llvm-svn: 169673
constructor/assignment operator with a const-qualified parameter type. The
prior method for determining this incorrectly used overload resolution.
llvm-svn: 168775
not ReadExpr().
Also add a test case making sure the thread safety attributes work as expected
when they come from a PCH.
Fixes rdar://12584141 & http://llvm.org/PR13982
llvm-svn: 168017
new container so we can safely iterate over them.
The container holding the lookup decls can under certain conditions
be changed while iterating (e.g. because of deserialization).
llvm-svn: 167816
reference instead of relying on computing it.
In general, if storage is no issue, it is preferable to deserialize info from
the PCH instead of trying to recompute it after the PCH was loaded.
The incentive to change this now was due to r155303 changing how friend template
classes in dependent contexts are handled; such classes can now be chained to
a previous template class but the computed InjectedClassNameType may be different
due to the extra template parameters from the dependent context.
The new handling requires more investigation but, in the meantime, writing out
InjectedClassNameType fixes PCH issue in rdar://12627738.
llvm-svn: 167425
The stat cache became essentially useless ever since we started
validating all file entries in the PCH.
But the motivating reason for removing it now is that it also affected
correctness in this situation:
-You have a header without include guards (using "#pragma once" or #import)
-When creating the PCH:
-The same header is referenced in an #include with different filename cases.
-In the PCH, of course, we record only one file entry for the header file
-But we cache in the PCH file the stat info for both filename cases
-Then the source files are updated and the header file is updated in a way that
its size and modification time are the same but its inode changes
-When using the PCH:
-We validate the headers, we check that header file and we create a file entry with its current inode
-There's another #include with a filename with different case than the previously created file entry
-In order to get its stat info we go through the cached stat info of the PCH and we receive the old inode
-because of the different inodes, we think they are different files so we go ahead and include its contents.
Removing the stat cache will potentially break clients that are attempting to use the stat cache
as a way of avoiding having the actual input files available. If that use case is important, patches are welcome
to bring it back in a way that will actually work correctly (i.e., emit a PCH that is self-contained, coping with
literal strings, line/column computations, etc.).
This fixes rdar://5502805
llvm-svn: 167172
diagnostic states; make sure the ASTReader sets the diagnostic state
properly instead of always recreating it.
Fixes rdar://12581618 & http://llvm.org/PR14181
llvm-svn: 166987
the macros that are #define'd or #undef'd on the command line. This
checking happens much earlier than the current macro-definition
checking and is far cleaner, because it does a direct comparison
rather than a diff of the predefines buffers. Moreover, it allows us
to use the result of this check to skip over PCH files within a
directory that have non-matching -D's or -U's on the command
line. Finally, it improves the diagnostics a bit for mismatches,
fixing <rdar://problem/8612222>.
The old predefines-buffer diff'ing will go away in a subsequent commit.
llvm-svn: 166641
check each of the files within that directory to determine if any of
them is an AST file that matches the language and target options. If
so, the first matching AST file is loaded. This fixes a longstanding
discrepency with GCC's precompiled header implementation.
llvm-svn: 166469
block, so the input files are validated early on, before we've
committed to loading the AST file. This (accidentally) fixed a but
wherein the main file used to generate the AST file would *not* be
validated by the existing validation logic.
At the moment, this leads to some duplication of filenames between the
source manager block and input-file blocks, as well as validation
logic. This will be handled via an upcoming patch.
llvm-svn: 166251
MacroInfo*. Instead of simply dumping an offset into the current file,
give each macro definition a proper ID with all of the standard
modules-remapping facilities. Additionally, when a macro is modified
in a subsequent AST file (e.g., #undef'ing a macro loaded from another
module or from a precompiled header), provide a macro update record
rather than rewriting the entire macro definition. This gives us
greater consistency with the way we handle declarations, and ties
together macro definitions much more cleanly.
Note that we're still not actually deserializing macro history (we
never were), but it's far easy to do properly now.
llvm-svn: 165560
Summary: Uses [[clang::fallthrough]] attribute in a template function, and -Wimplicit-fallthrough to check the attribute presence in an instantiation.
Reviewers: rsmith
Reviewed By: rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D55
llvm-svn: 165068
load in the IndirectField declarations as well.
Field designators in initializer lists depend on traversing the fields
decl chain to find the indirect fields.
Fixes rdar://12239321
llvm-svn: 163552
unexpanded parameter pack is a pack expansion. Thus, as with a non-type template
parameter which is a pack expansion, it needs to be expanded early into a fixed
list of template parameters.
Since the expanded list of template parameters is not itself a parameter pack,
it is permitted to appear before the end of the template parameter list, so also
remove that restriction (for both template template parameter pack expansions and
non-type template parameter pack expansions).
llvm-svn: 163369
This is accomplished by making VerifyDiagnosticsConsumer a CommentHandler,
which then only reads the -verify directives that are actually in live
blocks of code. It also makes it simpler to handle -verify directives that
appear in header files, though we still have to manually reparse some files
depending on how they are generated.
This requires some test changes. In particular, all PCH tests now have their
-verify directives outside the "header" portion of the file, using the @line
syntax added in r159978. Other tests have been modified mostly to make it
clear what is being tested, and to prevent polluting the expected output with
the directives themselves.
Patch by Andy Gibbs! (with slight modifications)
The new Frontend/verify-* tests exercise the functionality of this commit,
as well as r159978, r159979, and r160053 (Andy's other -verify enhancements).
llvm-svn: 160068
coming from an AST file are registered for serialization.
A static data member instantiation of in a chained PCH could be missed
when serializing decls; the result was that when emitting the visible decls
map of its DeclContext, we would use a DeclID that was not actually emitted,
leading to crashes or hangs.
Fix this by making sure such decls are always registered for serialization.
Also introduce extra sanity checks to make sure we don't register new
declarations or types after we have serialized the types/decls block.
rdar://11728990
llvm-svn: 159550
For some targets a structure named __va_list_tag is built to help define
the __builtin_va_list type. However, __va_list_tag was not being treated as a
predefined type thus causing problems when serializing the AST. This commit
fixes that oversight by adding the necessary support to treat __va_list_tag
as a predefined type.
llvm-svn: 159508
Also add a couple of unit tests to check the invalid-PCH error messages
to satisfy PR4568 and for the assertion (introduced in r149918 and fixed
in r158769) that would cause clang to crash when given an empty PCH.
llvm-svn: 158772
In standard C since C89, a 'translation-unit' is syntactically defined to have
at least one "external-declaration", which is either a decl or a function
definition. In Clang the latter gives us a declaration as well.
The tricky bit about this warning is that our predefines can contain external
declarations (__builtin_va_list and the 128-bit integer types). Therefore our
AST parser now makes sure we have at least one declaration that doesn't come
from the predefines buffer.
Also, remove bogus warning about empty source files. This doesn't catch source
files that only contain comments, and never fired anyway because of our
predefines.
PR12665 and <rdar://problem/9165548>
llvm-svn: 158085
Moves the bool bail-out down a little in SemaChecking - so now
-Wnull-conversion and -Wliteral-conversion can fire when the target type is
bool.
Also improve the wording/details in the -Wliteral-conversion warning to match
the -Wconstant-conversion.
llvm-svn: 156826
validate that we didn't override the contents of any of such files.
If this is detected, emit a diagnostic error and recover gracefully
by using the contents of the original file that the PCH was built from.
Part of rdar://11305263
llvm-svn: 156107
pretend there was no previous declaration -- that can lead us to injecting
a class template (with no access specifier) into a class scope. Instead,
just avoid the problematic checks.
llvm-svn: 155303
includes a patch from Matthias Kleine with a regression testcase!
Adds a new iterator 'data_iterator' to OnDiskHashTable which doesn't try to
reconstruct the external_key from the internal_key, which is useful for traits
that don't store enough information to do that mapping in their key. Also
deletes the 'item_iterator' from OnDiskHashTable as dead code.
llvm-svn: 154784
The warning this inhibits, -Wobjc-root-class, is opt-in for now. However, all clang unit tests that would trigger
the warning have been updated to use -Wno-objc-root-class. <rdar://problem/7446698>
llvm-svn: 154187
This diagnostic seems to be production ready, it's just an oversight that it
wasn't turned on by default.
The test changes are a bit of a mixed bag. Some tests that seemed like they
clearly didn't need to use this behavior have been modified not to use it.
Others that I couldn't be sure about, I added the necessary expected-warnings
to.
It's possible the diagnostic message could be improved to make it clearer that
this warning can be suppressed by using a value that won't lose precision when
converted to the target type (but can still be a floating point literal, such
as "bool b = 1.0;").
llvm-svn: 154068
InjectedClassNameType; otherwise, it won't be properly wired to the
original (canonical) declaration when it is deserialized. Fixes
<rdar://problem/11112464>.
llvm-svn: 153442
scoped enumeration members. Later uses of an enumeration temploid as a nested
name specifier should cause its instantiation. Plus some groundwork for
explicit specialization of member enumerations of class templates.
llvm-svn: 152750
analysis to make the AST representation testable. They are represented by a
new UserDefinedLiteral AST node, which is a sugared CallExpr. All semantic
properties, including full CodeGen support, are achieved for free by this
representation.
UserDefinedLiterals can never be dependent, so no custom instantiation
behavior is required. They are mangled as if they were direct calls to the
underlying literal operator. This matches g++'s apparent behavior (but not its
actual mangling, which is broken for literal-operator-ids).
User-defined *string* literals are now fully-operational, but the semantic
analysis is quite hacky and needs more work. No other forms of user-defined
literal are created yet, but the AST support for them is present.
This patch committed after midnight because we had already hit the quota for
new kinds of literal yesterday.
llvm-svn: 152211
NSNumber, and boolean literals. This includes both Sema and Codegen support.
Included is also support for new Objective-C container subscripting.
My apologies for the large patch. It was very difficult to break apart.
The patch introduces changes to the driver as well to cause clang to link
in additional runtime support when needed to support the new language features.
Docs are forthcoming to document the implementation and behavior of these features.
llvm-svn: 152137
data members for deleted or user-provided destructors.
Now it's computed in advance, serialize it, and in passing fix all the other
record DefinitionData flags whose serialization was missing.
llvm-svn: 151441
that provides the behavior of the C++11 library trait
std::is_trivially_constructible<T, Args...>, which can't be
implemented purely as a library.
Since __is_trivially_constructible can have zero or more arguments, I
needed to add Yet Another Type Trait Expression Class, this one
handling arbitrary arguments. The next step will be to migrate
UnaryTypeTrait and BinaryTypeTrait over to this new, more general
TypeTrait class.
Fixes the Clang side of <rdar://problem/10895483> / PR12038.
llvm-svn: 151352
block pointer that returns a block literal which captures (by copy)
the lambda closure itself. Some aspects of the block literal are left
unspecified, namely the capture variable (which doesn't actually
exist) and the body (which will be filled in by IRgen because it can't
be written as an AST).
Because we're switching to this model, this patch also eliminates
tracking the copy-initialization expression for the block capture of
the conversion function, since that information is now embedded in the
synthesized block literal. -1 side tables FTW.
llvm-svn: 151131
pointers and block pointers). We use dummy definitions to keep the
invariant that an implicit, used definition has a body; IR generation
will substitute the actual contents, since they can't be represented
as C++.
For the block pointer case, compute the copy-initialization needed to
capture the lambda object in the block, which IR generation will need
later.
llvm-svn: 150645
to pretty-print such function types better, and to fix a case where we were not
instantiating templates in lexical order. In passing, move the Variadic bit from
Type's bitfields to FunctionProtoType to get the Type bitfields down to 32 bits.
Also ensure that we always substitute the return type of a function when
substituting explicitly-specified arguments, since that can cause us to bail
out with a SFINAE error before we hit a hard error in parameter substitution.
llvm-svn: 150241
members of class templates so that their values can be used in ICEs. This
required reverting r105465, to get such instantiated members to be included in
serialized ASTs.
llvm-svn: 147023
variable is initialized by a non-constant expression, and pass in the variable
being declared so that earlier-initialized fields' values can be used.
Rearrange VarDecl init evaluation to make this possible, and in so doing fix a
long-standing issue in our C++ constant expression handling, where we would
mishandle cases like:
extern const int a;
const int n = a;
const int a = 5;
int arr[n];
Here, n is not initialized by a constant expression, so can't be used in an ICE,
even though the initialization expression would be an ICE if it appeared later
in the TU. This requires computing whether the initializer is an ICE eagerly,
and saving that information in PCH files.
llvm-svn: 146856
chains. The previous implementation relied heavily on the declaration
chain being stored as a (circular) linked list on disk, as it is in
memory. However, when deserializing from multiple modules, the
different chains could get mixed up, leading to broken declaration chains.
The new solution keeps track of the first and last declarations in the
chain for each module file. When we load a declaration, we search all
of the module files for redeclarations of that declaration, then
splice together all of the lists into a coherent whole (along with any
redeclarations that were actually parsed).
As a drive-by fix, (de-)serialize the redeclaration chains of
TypedefNameDecls, which had somehow gotten missed previously. Add a
test of this serialization.
This new scheme creates a redeclaration table that is fairly large in
the PCH file (on the order of 400k for Cocoa.h's 12MB PCH file). The
table is mmap'd in and searched via a binary search, but it's still
quite large. A future tweak will eliminate entries for declarations
that have no redeclarations anywhere, and should
drastically reduce the size of this table.
llvm-svn: 146841
diagnostic message are compared. If either is a substring of the other, then
no error is given. This gives rise to an unexpected case:
// expect-error{{candidate function has different number of parameters}}
will match the following error messages from Clang:
candidate function has different number of parameters (expected 1 but has 2)
candidate function has different number of parameters
It will also match these other error messages:
candidate function
function has different number of parameters
number of parameters
This patch will change so that the verification string must be a substring of
the diagnostic message before accepting. Also, all the failing tests from this
change have been corrected. Some stats from this cleanup:
87 - removed extra spaces around verification strings
70 - wording updates to diagnostics
40 - extra leading or trailing characters (typos, unmatched parens or quotes)
35 - diagnostic level was included (error:, warning:, or note:)
18 - flag name put in the warning (-Wprotocol)
llvm-svn: 146619
we may end up having added more pending stuff to do, so go in a loop until everything
is cleared out.
This fixes the error in rdar://10278815 which has a certain David Lynch-esque quality..
error: unknown type name 'BOOL'; did you mean 'BOOL'?
llvm-svn: 145536
it is going to be rewritten (and the chain will be serialized again), otherwise we may form a cycle in its
categories list when deserializing.
Also introduce ASTMutationListener::CompletedObjCForwardRef to notify that a forward reference
was completed; using Decl's isChangedSinceDeserialization/setChangedSinceDeserialization
is bug inducing and kinda gross, we should phase it out.
Fixes infinite loop in rdar://10418538.
llvm-svn: 144465
path. That assumption should never have been true, but it was until
I fixed it. Now that its fixed, add a triple here to get correct
behavior even on Windows.
llvm-svn: 143863
to do "realistic" includes, and so need the header search logic now in
the driver. This in turn requires switching the CC1 options to the
actual driver options, and passing -Xclang where there is no analogy.
llvm-svn: 143805
statements. As noted in the documentation for the AST node, the
semantics of __if_exists/__if_not_exists are somewhat different from
the way Visual C++ implements them, because our parsed-template
representation can't accommodate VC++ semantics without serious
contortions. Hopefully this implementation is "good enough".
llvm-svn: 142901
in such a case just write out a reference of a previously serialized Stmt, instead
of serializing it all over again.
This saves memory + space + [de]serializing time, and avoids blowing up memory
with pathological cases. rdar://10293911
llvm-svn: 142696
increasingly prevailing case to the point that new features
like ARC don't even support the fragile ABI anymore.
This required a little bit of reshuffling with exceptions
because a check was assuming that ObjCNonFragileABI was
only being set in ObjC mode, and that's actually a bit
obnoxious to do.
Most, though, it involved a perl script to translate a ton
of test cases.
Mostly no functionality change for driver users, although
there are corner cases with disabling language-specific
exceptions that we should handle more correctly now.
llvm-svn: 140957
check whether the requested location points inside the precompiled preamble,
in which case the returned source location will be a "loaded" one.
llvm-svn: 140060
feature akin to the ARC runtime checks. Removes a terrible hack where
IR gen needed to find the declarations of those symbols in the translation
unit.
llvm-svn: 139404
instead of codegen waiting to consume such a declaration, which won't
happen if that decls are coming from a PCH.
Fixes rdar://10028656.
llvm-svn: 139359
'id' that can be used (only!) via a contextual keyword as the result
type of an Objective-C message send. 'instancetype' then gives the
method a related result type, which we have already been inferring for
a variety of methods (new, alloc, init, self, retain). Addresses
<rdar://problem/9267640>.
llvm-svn: 139275
The initial incentive was to fix a crash when PCH chaining categories
to an interface, but the fix was done in the "modules way" that I hear
is popular with the kids these days.
Each module stores the local chain of categories and we combine them
when the interface is loaded. We also warn if non-dependent modules
introduce duplicate named categories.
llvm-svn: 138926
sure that all of the CXXConversionDecls go into the same
bucket. Otherwise, name lookup might not find them all. Fixes
<rdar://problem/10041960>.
llvm-svn: 138824
redeclarations of a particular entity would occur in source
order. Friend declarations that occur within class templates (or
member classes thereof) do not follow this, nor would modules. Big
thanks to Erik Verbruggen for reducing this problem from the Very
Large Qt preamble testcase he found.
llvm-svn: 138557
by eliminating the type ID from constructor, destructor, and
conversion function names. There are several reasons for this change:
- A given type (say, int*) isn't guaranteed to have a single, unique
type ID within a chain of PCH files. Hence, we could end up hashing
based on the wrong type ID, causing name lookup to fail.
- The mapping from types back to type IDs required one DenseMap
entry for every type that was ever deserialized, which was an
unacceptable cost to support just the name lookup of constructors,
destructors, and conversion functions. Plus, this mapping could
never actually work with chained or multiple PCH, based on the first
bullet.
Once we have eliminated the type from the hash function, these
problems go away, as does my horrible "reverse type remap" hack, which
was doomed from the start (see bullet #1 above) and far too
complicated.
However, note that removing the type from the hash function means that
all constructors, destructors, and conversion functions have the same
hash key, so I've updated the caller to double-check that the
declarations found have the appropriate name.
llvm-svn: 136708
source locations from source locations loaded from an AST/PCH file.
Previously, loading an AST/PCH file involved carefully pre-allocating
space at the beginning of the source manager for the source locations
and FileIDs that correspond to the prefix, and then appending the
source locations/FileIDs used for parsing the remaining translation
unit. This design forced us into loading PCH files early, as a prefix,
whic has become a rather significant limitation.
This patch splits the SourceManager space into two parts: for source
location "addresses", the lower values (growing upward) are used to
describe parsed code, while upper values (growing downward) are used
for source locations loaded from AST/PCH files. Similarly, positive
FileIDs are used to describe parsed code while negative FileIDs are
used to file/macro locations loaded from AST/PCH files. As a result,
we can load PCH/AST files even during parsing, making various
improvemnts in the future possible, e.g., teaching #include <foo.h> to
look for and load <foo.h.gch> if it happens to be already available.
This patch was originally written by Sebastian Redl, then brought
forward to the modern age by Jonathan Turner, and finally
polished/finished by me to be committed.
llvm-svn: 135484
When a macro instantiation occurs, reserve a SLocEntry chunk with length the
full length of the macro definition source. Set the spelling location of this chunk
to point to the start of the macro definition and any tokens that are lexed directly
from the macro definition will get a location from this chunk with the appropriate offset.
For any tokens that come from argument expansion, '##' paste operator, etc. have their
instantiation location point at the appropriate place in the instantiated macro definition
(the argument identifier and the '##' token respectively).
This improves macro instantiation diagnostics:
Before:
t.c:5:9: error: invalid operands to binary expression ('struct S' and 'int')
int y = M(/);
^~~~
t.c:5:11: note: instantiated from:
int y = M(/);
^
After:
t.c:5:9: error: invalid operands to binary expression ('struct S' and 'int')
int y = M(/);
^~~~
t.c:3:20: note: instantiated from:
\#define M(op) (foo op 3);
~~~ ^ ~
t.c:5:11: note: instantiated from:
int y = M(/);
^
The memory savings for a candidate boost library that abuses the preprocessor are:
- 32% less SLocEntries (37M -> 25M)
- 30% reduction in PCH file size (900M -> 635M)
- 50% reduction in memory usage for the SLocEntry table (1.6G -> 800M)
llvm-svn: 134587
vector<int>
to
std::vector<int>
Patch by Kaelyn Uhrain, with minor tweaks + PCH support from me. Fixes
PR5776/<rdar://problem/8652971>.
Thanks Kaelyn!
llvm-svn: 134007
Language-design credit goes to a lot of people, but I particularly want
to single out Blaine Garst and Patrick Beard for their contributions.
Compiler implementation credit goes to Argyrios, Doug, Fariborz, and myself,
in no particular order.
llvm-svn: 133103
Related result types apply Cocoa conventions to the type of message
sends and property accesses to Objective-C methods that are known to
always return objects whose type is the same as the type of the
receiving class (or a subclass thereof), such as +alloc and
-init. This tightens up static type safety for Objective-C, so that we
now diagnose mistakes like this:
t.m:4:10: warning: incompatible pointer types initializing 'NSSet *'
with an
expression of type 'NSArray *' [-Wincompatible-pointer-types]
NSSet *array = [[NSArray alloc] init];
^ ~~~~~~~~~~~~~~~~~~~~~~
/System/Library/Frameworks/Foundation.framework/Headers/NSObject.h:72:1:
note:
instance method 'init' is assumed to return an instance of its
receiver
type ('NSArray *')
- (id)init;
^
It also means that we get decent type inference when writing code in
Objective-C++0x:
auto array = [[NSMutableArray alloc] initWithObjects:@"one", @"two",nil];
// ^ now infers NSMutableArray* rather than id
llvm-svn: 132868
a file was modified since the time the PCH was created.
The parser is not fit to deal with stale PCHs, too many invariants do not hold up. rdar://9530587.
llvm-svn: 132389
Increase robustness of the delegating constructor cycle detection
mechanism. No more infinite loops on invalid or logic errors leading to
false results. Ensure that this is maintained correctly accross
serialization.
llvm-svn: 130887
This is more efficient as it's all done at once at the end of the TU.
This could still get expensive, so a flag is provided to disable it. As
an added bonus, the diagnostics will now print out a cycle.
The PCH test is XFAILed because we currently can't deal with a note
emitted in the header and I, being tired, see no other way to verify the
serialization of delegating constructors. We should probably address
this problem /somehow/ but no good solution comes to mind.
llvm-svn: 130836
required modifying a few tests that specifically use note include stacks
to check the source manager's view of include stacks. I've simply added
the flag to these tests for now, they may have to be more substantially
changed if we decide to remove support for note include stacks
altogether.
Also, add a test for include stacks on notes that was supposed to go in
with the previous commit.
llvm-svn: 128390
This is basically the same idea as the warning on uninitialized uses of
fields within an initializer list. As such, it is on by default and
under -Wuninitialized.
Original patch by Richard Trieu, with some massaging from me on the
wording and grouping of the diagnostics.
llvm-svn: 128376
arguments at the same offset, since it's needed when creating the empty
DeclRefExpr when deserializing. Fixes a memory corruption issue that would lead
to random bugs and crashes.
llvm-svn: 127125
use the translation unit as its declaration context, then deserialize
the actual lexical and semantic DeclContexts after the template
parameter is complete. This avoids problems when the DeclContext
itself (e.g., a class template) is dependent on the template parameter
(e.g., for the injected-class-name).
llvm-svn: 127056
Store in PCH the directory that the PCH was originally created in.
If a header file is not found at the path that we expect it to be and the PCH file
was moved from its original location, try to resolve the file by assuming that
header+PCH were moved together and the header is in the same place relative to the PCH.
llvm-svn: 125576
we would deserialize all of the macro definitions we knew about while
serializing the macro definitions at the end of the AST/PCH file. Even
though we skipped most of them (since they were unchanged), it's still
a performance problem.
Now, we do the standard AST/PCH chaining trick: watch what identifiers
are deserialized as macro names, and consider only those identifiers
(along with macro definitions that have been deserialized/written in
the source) when serializing the preprocessor state.
llvm-svn: 125324
FileManager.cpp: Allow virtual files in nonexistent directories.
FileManager.cpp: Close FileDescriptor for virtual files that correspond to actual files.
FileManager.cpp: Enable virtual files to be created even for files that were flagged as NON_EXISTENT_FILE, e.g. by a prior (unsuccessful) addFile().
ASTReader.cpp: Read a PCH even if the original source files cannot be found.
Add a test for reading a PCH of a file that has been removed and diagnostics referencing that file.
llvm-svn: 124374
getCanonicalType() to make sure that the type we got back is actually
canonical. This is the case for most types, which always build a
canonical type when given canonical components. However, some types that
involve expressions in their canonicalization (e.g., array types with
dependent sizes) don't always build canonical types from canonical
components, because there is no such thing as a "canonical"
expression. Therefore, we do this extra mapping to ensure that the
canonical types we store are actually canonical.
llvm-svn: 117344
In that case a chained PCH will record the updates to the DefinitionData pointer of forward references.
If a forward reference mutated into a definition re-write it into the chained PCH, this is too big of a change.
llvm-svn: 117239
identifiers to determine good typo-correction candidates. Once we've
identified those candidates, we perform name lookup on each of them
and the consider the results.
This optimization makes typo correction > 2x faster on a benchmark
example using a single typo (NSstring) in a tiny file that includes
Cocoa.h from a precompiled header, since we are deserializing far less
information now during typo correction.
There is a semantic change here, which is interesting. The presence of
a similarly-named entity that is not visible can now affect typo
correction. This is both good (you won't get weird corrections if the
thing you wanted isn't in scope) and bad (you won't get good
corrections if there is a similarly-named-but-completely-unrelated
thing). Time will tell whether it was a good choice or not.
llvm-svn: 116528
instead of deserializing the complete declaration context of the record.
Iterating over the fields of a record is very common (e.g to determine the layout), unfortunately we needlessly deserialize every declaration
that the declaration context of the record contains; this can be bad for large C++ classes that contain a lot of methods.
Fix this by allow deserialization of just the fields when we want to iterate over them.
Progress for rdar://7260160.
llvm-svn: 116507
following amusing sequence:
- AST writing schedules writing a type X* that it had never seen
before
- AST writing starts writing another declaration, ends up
deserializing X* from a prior AST file. Now we have two type IDs for
the same type!
- AST writer tries to write X*. It only has the lower-numbered ID
from the the prior AST file, so references to the higher-numbered ID
that was scheduled for writing go off into lalaland.
To fix this, keep the higher-numbered ID so we end up writing the type
twice. Since this issue occurs so rarely, and type records are
generally rather small, I deemed this better than the alternative: to
keep a separate mapping from the higher-numbered IDs to the
lower-numbered IDs, which we would end up having to check whenever we
want to deserialize any type.
Fixes <rdar://problem/8511624>, I think.
llvm-svn: 115647
waiting until we think we need it: we didn't catch all of the places
where we actually needed it, and we probably wouldn't ever. Fixes a
C++ PCH crasher.
llvm-svn: 115617
the interface as having changed since it was originally
serialized. This ensures that we see class extensions/categories in
chained PCH files.
llvm-svn: 115421
Guy Benyei