Serge Pavlov [Wed, 14 Jun 2017 10:07:02 +0000 (10:07 +0000)]
Function with unparsed body is a definition
While a function body is being parsed, the function declaration is not considered
as a definition because it does not have a body yet. In some cases it leads to
incorrect interpretation, the case is presented in
https://bugs.llvm.org/show_bug.cgi?id=14785:
```
template<typename T> struct Somewhat {
void internal() const {}
friend void operator+(int const &, Somewhat<T> const &) {}
};
void operator+(int const &, Somewhat<char> const &x) { x.internal(); }
```
When statement `x.internal()` in the body of global `operator+` is parsed, the type
of `x` must be completed, so the instantiation of `Somewhat<char>` is started. It
instantiates the declaration of `operator+` defined inline, and makes a check for
redefinition. The check does not detect another definition because the declaration
of `operator+` is still not defining as does not have a body yet.
To solves this problem the function `isThisDeclarationADefinition` considers
a function declaration as a definition if it has flag `WillHaveBody` set.
Eric Fiselier [Wed, 14 Jun 2017 03:24:55 +0000 (03:24 +0000)]
[coroutines] Fix co_await for range statement
Summary:
Currently we build the co_await expressions on the wrong implicit statements of the implicit ranged for; Specifically we build the co_await expression wrapping the range declaration, but it should wrap the begin expression.
Florian Hahn [Tue, 13 Jun 2017 18:06:15 +0000 (18:06 +0000)]
Align definition of DW_OP_plus with DWARF spec [2/3]
Summary:
This patch is part of 3 patches that together form a single patch, but must be introduced in stages in order not to break things.
The way that LLVM interprets DW_OP_plus in DIExpression nodes is basically that of the DW_OP_plus_uconst operator since LLVM expects an unsigned constant operand. This unnecessarily restricts the DW_OP_plus operator, preventing it from being used to describe the evaluation of runtime values on the expression stack. These patches try to align the semantics of DW_OP_plus and DW_OP_minus with that of the DWARF definition, which pops two elements off the expression stack, performs the operation and pushes the result back on the stack.
This is done in three stages:
• The first patch (LLVM) adds support for DW_OP_plus_uconst and changes all uses (and tests) of DW_OP_plus to use DW_OP_plus_uconst.
• The second patch (Clang) contains changes to use DW_OP_plus_uconst instead of DW_OP_plus.
• The third patch (LLVM) changes the semantics of DW_OP_plus to be in line with it’s DWARF meaning. It also does this for DW_OP_minus.
Francois Ferrand [Tue, 13 Jun 2017 07:02:43 +0000 (07:02 +0000)]
clang-format: add option to merge empty function body
Summary:
This option supplements the AllowShortFunctionsOnASingleLine flag, to
merge empty function body at the beginning of the line: e.g. when the
function is not short-enough and breaking braces after function.
Reid Kleckner [Mon, 12 Jun 2017 19:57:56 +0000 (19:57 +0000)]
Correct debug info bit offset calculation for big-endian targets
Summary:
The change "[CodeView] Implement support for bit fields in
Clang" (r274201, https://reviews.llvm.org/rL274201) broke the
calculation of bit offsets for the debug info describing bitfields on
big-endian targets.
Prior to commit r274201 the debug info for bitfields got their offsets
from the ASTRecordLayout in CGDebugInfo::CollectRecordFields(), the
current field offset was then passed on to
CGDebugInfo::CollectRecordNormalField() and used directly in the
DIDerivedType.
Since commit r274201, the bit offset ending up in the DIDerivedType no
longer comes directly from the ASTRecordLayout. Instead
CGDebugInfo::CollectRecordNormalField() calls the new method
CGDebugInfo::createBitFieldType(), which in turn calls
CodeGenTypes::getCGRecordLayout().getBitFieldInfo() to fetch a
CGBitFieldInfo describing the field. The 'Offset' member of
CGBitFieldInfo is then used to calculate the bit offset of the
DIDerivedType. Unfortunately the previous and current method of
calculating the bit offset are only equivalent for little endian
targets, as CGRecordLowering::setBitFieldInfo() reverses the bit
offsets for big endian targets as the last thing it does.
A simple reproducer for this error is the following module:
struct fields {
unsigned a : 4;
unsigned b : 4;
} flags = {0x0f, 0x1};
Compiled for Mips, with commit r274200 both the DIDerivedType bit
offsets on the IR-level and the DWARF information on the ELF-level
will have the expected values: the offsets of 'a' and 'b' are 0 and 4
respectively. With r274201 the offsets are switched to 4 and 0. By
noting that the static initialization of 'flags' in both cases is the
same, we can eliminate a change in record layout as the cause of the
change in the debug info. Also compiling this example with gcc,
produces the same record layout and debug info as commit r274200.
In order to restore the previous function we extend
CGDebugInfo::createBitFieldType() to compensate for the reversal done
in CGRecordLowering::setBitFieldInfo().
Vedant Kumar [Mon, 12 Jun 2017 18:42:51 +0000 (18:42 +0000)]
[ubsan] Detect invalid unsigned pointer index expression (clang)
Adding an unsigned offset to a base pointer has undefined behavior if
the result of the expression would precede the base. An example from
@regehr:
int foo(char *p, unsigned offset) {
return p + offset >= p; // This may be optimized to '1'.
}
foo(p, -1); // UB.
This patch extends the pointer overflow check in ubsan to detect invalid
unsigned pointer index expressions. It changes the instrumentation to
only permit non-negative offsets in pointer index expressions when all
of the GEP indices are unsigned.
Testing: check-llvm, check-clang run on a stage2, ubsan-instrumented
build.
Artem Dergachev [Mon, 12 Jun 2017 17:59:50 +0000 (17:59 +0000)]
[analyzer] Fix a crash when an ObjC object is constructed in AllocaRegion.
Memory region allocated by alloca() carries no implicit type information.
Don't crash when resolving the init message for an Objective-C object
that is being constructed in such region.
These options control the behaviour of the compression of debug info
sections on ELF targets. Our behaviour slightly diverges from the
behaviour of GCC. `-gz` maps to the `-compress-debug-sections` rather
than `-compress-debug-sections=zlib` or
`-compress-debug-sections=zlib-gnu`. This small divergence allows us to
be compatible across versions of binutils (=zlib support was introduced
in 2.26, while earlier versions only support =zlib-gnu). This also
allows users to not have to worry about the version of the assembler
they may be using if they are not using the IAS. Previously, users
would have had to go through the internal option
`-compress-debug-sectionss` and pass that through to the assembler,
which is no longer needed.
Driver: pass along [-]-[no]compress-debug-sections unfiltered
Rather than validating the flags, pass them through without any
validation. Arguments passed via -Wa or -Xassembler are passed directly
to the assembler without validation. The validation was previously
required since we did not provide proper driver level support for
controlling the debug compression on ELF targets. A subsequent change
will add support for the `-gz` and `-gz=` flags which provide proper
driver level control of the ELF compressed debug sections.
Erich Keane [Fri, 9 Jun 2017 22:50:02 +0000 (22:50 +0000)]
Support operator keywords used in Windows SDK(fix ubsan)
UBSan found an issue with a nullptr being assigned to a reference.
This was because a following function went back and checked the
identifier in the CPPOperatorName case. This patch corrects that
location with the original logic as well.
Vassil Vassilev [Fri, 9 Jun 2017 21:36:28 +0000 (21:36 +0000)]
[modules] Fix that global delete operator get's assigned to a submodule.
n the current local-submodule-visibility mode, as soon as we discover a virtual
destructor, we declare on demand a global delete operator. However, this causes
that this delete operator is owned by the submodule which contains said virtual
destructor. This means that other modules no longer can see the global delete
operator which is hidden inside another submodule and fail to compile.
This patch unhides those global allocation function once they're created to
prevent this issue.
Richard Smith [Fri, 9 Jun 2017 21:24:02 +0000 (21:24 +0000)]
Add -frewrite-imports flag.
If specified, when preprocessing, the contents of imported .pcm files will be
included in preprocessed output. The resulting preprocessed file can then be
compiled standalone without the module sources or .pcm files.
Richard Smith [Fri, 9 Jun 2017 19:22:32 +0000 (19:22 +0000)]
Add #pragma clang module build/endbuild pragmas for performing a module build
as part of a compilation.
This is intended for two purposes:
1) Writing self-contained test cases for modules: we can now write a single
source file test that builds some number of module files on the side and
imports them.
2) Debugging / test case reduction. A single-source testcase is much more
amenable to reduction, compared to a VFS tarball or .pcm files.
Vassil Vassilev [Fri, 9 Jun 2017 16:42:26 +0000 (16:42 +0000)]
Repair 2010-05-31-palignr.c test
This test was silently failing since a long time because it failed to include
stdlib.h (as it's running in a freestanding environment). However, because we
used just not clang_cc1 instead of the verify mode, this regression was never
noticed and the test was just always passing.
This adds -ffreestanding to the invocation, so that tmmintrin.h doesn't
indirectly include mm_malloc.h, which in turns includes the unavailable stdlib.h.
We also run now in the -verify mode to prevent that we silently regress again.
I've also updated the test to no longer check the return value of _mm_alignr_epi8
as this is also causing it to fail (and it's not really the job of this test to
test this).
Erich Keane [Fri, 9 Jun 2017 16:29:35 +0000 (16:29 +0000)]
support operator keywords used in Windows SDK
to support operator keywords used in Windows SDK, alter token type when
seen in system headers
Hello, I submitted D33505 to address this problem, but the
proposal was rejected as too big a hammer.
This change will allow clang to parse the WindowsSDK header <query.h>
which uses the operator name "or" as a field name. Treat cpp operator
keywords as ordinary identifiers inside the Microsoft headers, but
treat them as usual in the user's program.
Alexey Bataev [Fri, 9 Jun 2017 13:40:18 +0000 (13:40 +0000)]
[DebugInfo] Add kind of ImplicitParamDecl for emission of FlagObjectPointer.
Summary:
If the first parameter of the function is the ImplicitParamDecl, codegen
automatically marks it as an implicit argument with `this` or `self`
pointer. Added internal kind of the ImplicitParamDecl to separate
'this', 'self', 'vtt' and other implicit parameters from other kind of
parameters.
Summary:
- Implements TargetInfo class for Nios2 target.
- Enables handling of -march and -mcpu options for Nios2 target.
- Definition of Nios2 builtin functions.
Erik Verbruggen [Fri, 9 Jun 2017 08:29:58 +0000 (08:29 +0000)]
Speed up preamble loading
Cache filename - SourceLocation pairs to speed up preamble loading and
global completion. This is especially relevant for windows, where
preamble loading takes a while.
Richard Smith [Fri, 9 Jun 2017 01:36:10 +0000 (01:36 +0000)]
Remove 'Filename' parameter from BeginSourceFileAction.
No-one was using this, and it's not meaningful in general -- FrontendActions
can be run on inputs that don't have a corresponding source file. The current
frontend input can be obtained by asking the FrontendAction if any future
action actually needs it.
[libclang] Introduce a new parsing option 'CXTranslationUnit_SingleFileParse' that puts preprocessor in a mode for parsing a single file only.
This is useful for parsing a single file, as a fast/inaccurate 'mode' that can still provide declarations from the file, like the classes and their methods.
Represent debug information compression type fully
This is tied with the LLVM side of the change to expose the debug
information compression types to clang. We now track the compression
type as an enumeration rather than a boolean. We still use the same
value (GNU) that we did previously. This is in preparation to support
passing down the compression type and switch it based on the command
line.
[sanitizer-coverage] one more flavor of coverage: -fsanitize-coverage=inline-8bit-counters. Experimental so far, not documenting yet. Reapplying revisions 304630, 304631, 304632, 304673, see PR33308
Peter Wu [Thu, 8 Jun 2017 22:58:12 +0000 (22:58 +0000)]
[ASTMatchers] temporary disable tests with floating suffix
r305022 assumed that floatLiteral(equals(1.2)) would also match 1.2f and
1.2l, but apparently that is not the case. Until it is clear how to
match, temporary disable the test to fix CI.
Peter Wu [Thu, 8 Jun 2017 22:00:58 +0000 (22:00 +0000)]
[ASTMatchers] Add clang-query support for equals matcher
Summary:
This allows the clang-query tool to use matchers like
"integerLiteral(equals(32))". For this to work, an overloaded function
is added for each possible parameter type.
Peter Wu [Thu, 8 Jun 2017 22:00:50 +0000 (22:00 +0000)]
[ASTMatchers] Add support for floatLiterals
Summary:
Needed to support something like "floatLiteral(equals(1.0))". The
parser for floating point numbers is kept simple, so instead of ".1" you
have to use "0.1".
Peter Wu [Thu, 8 Jun 2017 22:00:38 +0000 (22:00 +0000)]
[ASTMatchers] Add support for boolean literals
Summary:
Recognize boolean literals for future extensions ("equals(true)").
Note that a specific VariantValue constructor is added to resolve
ambiguity (like "Value = 5") between unsigned and bool.
This diff fixes printf "fixits" in the case when there is
a wrapping macro and the format string needs multiple replacements.
In the presence of a macro there is an extra logic in EditedSource.cpp
to handle multiple uses of the same macro argument
(see the old comment inside EditedSource::canInsertInOffset)
which was mistriggerred when the argument was used only once
but required multiple adjustments), as a result the "fixit"
was breaking down the format string
by dropping the second format specifier, i.e.
Log1("test 4: %s %s", getNSInteger(), getNSInteger())
was getting replaced with
Log1("test 4: %ld ", (long)getNSInteger(), (long)getNSInteger())
(if one removed the macro and used printf directly it would work fine).
In this diff we track the location where the macro argument is used and
(as it was before) the modifications originating from all the locations
except the first one are rejected, but multiple changes are allowed.
[Sema] Remove unused field from OverloadCandidate.
The only use in-tree I can find for BuiltinTypes.ResultTy is a single
store to it. We otherwise just recompute what it should be later on (and
sometimes do things like argument conversions in the process of
recomputing it).
Since it's impossible to test if the value stored there is sane, and we
don't use it anyway, we should probably just drop the field.
I'll do a follow-up patch to rename BuiltinTypes.ParamTypes ->
BuiltinParamTypes in a bit. Wanted to keep this patch relatively
minimal.
Summary:
- Implements TargetInfo class for Nios2 target.
- Enables handling of -march and -mcpu options for Nios2 target.
- Definition of Nios2 builtin functions.
Serge Pavlov [Thu, 8 Jun 2017 06:31:19 +0000 (06:31 +0000)]
Do not inherit default arguments for friend function in class template.
A function declared in a friend declaration may have declarations prior
to the containing class definition. If such declaration defines default
argument, the friend function declaration inherits them. This behavior
causes problems if the class where the friend is declared is a template:
during the class instantiation the friend function looks like if it had
default arguments, so error is triggered.
With this change friend functions declared in class templates do not
inherit default arguments. Actual set of them will be defined at the
point where the containing class is instantiated.
Serge Pavlov [Thu, 8 Jun 2017 06:07:07 +0000 (06:07 +0000)]
Improve diagnostics if friend function redefines file-level function.
Clang makes check for function redefinition after it merged the new
declaration with the existing one. As a result, it produces poor
diagnostics in the case of a friend function defined inline, as in
the code:
```
void func() {}
class C { friend void func() {} };
```
Error message in this case states that `inline declaration of 'func'
follows non-inline definition`, which is misleading, as `func` does
not have explicit `inline` specifier.
With this changes compiler reports function redefinition if the new
function is a friend defined inline and it does not have explicit
`inline` specifier.
Serge Pavlov [Thu, 8 Jun 2017 05:25:19 +0000 (05:25 +0000)]
Catch invalid bitwise operation on vector of floats
Bitwise complement applied to vector of floats described with
attribute `ext_vector_type` is not diagnosed as error. Attempt to
compile such construct causes assertion violation in Instruction.cpp.
With this change the complement is treated similar to the case of
vector type described with attribute `vector_size`.
Richard Smith [Thu, 8 Jun 2017 01:08:50 +0000 (01:08 +0000)]
Weaken restriction in r304862 to allow implicit deduction guides to reference
the injected-class-name of a specialization that uses a partial / explicit
specialization.
Petar Jovanovic [Wed, 7 Jun 2017 23:51:52 +0000 (23:51 +0000)]
Reapply r304929 [mips] Add runtime options to enable/disable madd/sub.fmt
The test in r304929 broke multiple buildbots as it expected mips target to
be registered and available (which is not necessarily true). Updating the
test with this condition.
Original commit:
[mips] Add runtime options to enable/disable madd.fmt and msub.fmt
Add options to clang: -mmadd4 and -mno-madd4, use it to enable or disable
generation of madd.fmt and similar instructions respectively, as per GCC.
Richard Smith [Wed, 7 Jun 2017 21:46:22 +0000 (21:46 +0000)]
[c++1z] Support deducing B in noexcept(B).
This is not required by the standard (yet), but there seems to be reasonable
support for this being a defect according to CWG discussion, and libstdc++ 7.1
relies on it working.
[clang-format] Fix alignment of preprocessor trailing comments
Summary:
This patch is a follow-up of https://reviews.llvm.org/rL304687, which fixed an
overflow in the comment alignment code in clang-format. The token length of
trailing comments of preprocessor directives is calculated incorrectly by
including the text between consecutive directives. That causes them to not being
aligned.
For example, in this code with column limit 20
```
#if A
#else // A
int iiii;
#endif // B
```
the length of the token `// A` was wrongly calculated as 14 = 5 (the size of `// A\n`) plus 9 (the size of `int iiii;`) and so `// A` wouldn't be aligned with `// B` and this was produced:
```
#if A
#else // A
int iiii;
#endif // B
```
Dimitry Andric [Wed, 7 Jun 2017 12:05:41 +0000 (12:05 +0000)]
Only print registered targets for `--version`
Summary:
In D33900, I added printing of the registered targets in clang's
`PrintVersion` function, which is not only used for `--version` output,
but also for `-v` (verbose mode) and `-###`. Especially the latter
seems to trip up some test cases, so it is probably better to only print
the registered targets for `--version`.
Florian Hahn [Wed, 7 Jun 2017 11:50:45 +0000 (11:50 +0000)]
[CodeGen] Add thumb-mode to target-features for arm/thumb triples.
Summary:
The thumb-mode target feature is used to force Thumb or ARM code
generation on a per-function basis. Explicitly adding +thumb-mode to
functions for thumbxx triples enables mixed ARM/Thumb code generation in
places where compilation units with thumbxx and armxx triples are merged
together (e.g. the IR linker or LTO).
For armxx triples, -thumb-mode is added in a similar fashion.
Javed Absar [Wed, 7 Jun 2017 10:02:02 +0000 (10:02 +0000)]
[ARM] Fix Neon vector type alignment to 64-bit
This is restricted version of patch - https://reviews.llvm.org/D33205
that I reverted as it was leading to ABI breaks on darwin etc.
This patch restricts the fix to AAPCS (Android remains 128-bit).
Reviewed by: Renato Golin, Stephen Hines
Differential Revision: https://reviews.llvm.org/D33786
Richard Smith [Wed, 7 Jun 2017 02:42:27 +0000 (02:42 +0000)]
Fix a couple of class template argument deduction crashes with libc++'s tuple.
RecursiveASTVisitor was not properly recursing through a
SubstTemplateTypeParmTypes, resulting in crashes in pack expansion where we
couldn't always find an unexpanded pack within a pack expansion.
We also have an issue where substitution of deduced template arguments for an
implicit deduction guide creates the "impossible" case of naming a
non-dependent member of the current instantiation, but within a specialization
that is actually instantiated from a different (partial/explicit)
specialization of the template. We resolve this by declaring that constructors
that do so can only be used to deduce specializations of the primary template.
I'm running this past CWG to see if people agree this is the right thing to do.
Richard Smith [Wed, 7 Jun 2017 00:29:44 +0000 (00:29 +0000)]
Improve error recovery for missing 'template' keyword in contexts where the
template is valid with or without it (with different meanings).
If we see "dependent.x<...", and what follows the '<' is a valid expression,
we must parse the '<' as a comparison rather than a template angle bracket.
When we later come to instantiate, if we find that the LHS of the '<' actually
names an overload set containing function templates, produce a diagnostic
suggesting that the 'template' keyword was missed rather than producing a
mysterious diagnostic saying that the function must be called (and pointing
at what looks to already be a function call!).
projects / clang / log