Alexey Bataev [Thu, 25 Oct 2018 15:35:27 +0000 (15:35 +0000)]
[OPENMP]Fix PR39422: variables are not firstprivatized in task context.
According to the OpenMP standard, In a task generating construct, if no
default clause is present, a variable for which the data-sharing
attribute is not determined by the rules above is firstprivatized.
Compiler tries to implement this, but if the variable is not directly
used in the task context, this variable may not be firstprivatized.
Patch fixes this problem.
Will Wilson [Thu, 25 Oct 2018 11:45:32 +0000 (11:45 +0000)]
[ms] Prevent explicit constructor name lookup if scope is missing
MicrosoftExt allows explicit constructor calls. Prevent lookup of constructor name unless the name has explicit scope.
This avoids a compile-time crash due to confusing a member access for a constructor name.
[clang-format] Break before next parameter after a formatted multiline raw string parameter
Summary:
Currently clang-format breaks before the next parameter after multiline parameters (also recursively for the parent expressions of multiline parameters). However, it fails to do so for formatted multiline raw string literals:
```
$ cat test.cc
// Examples
// Regular multiline tokens
int x = f(R"(multi
line)", 2);
}
int y = g(h(R"(multi
line)"), 2);
// Formatted multiline tokens
int z = f(R"pb(multi: 1 #
line: 2)pb", 2);
int w = g(h(R"pb(multi: 1 #
line: 2)pb"), 2);
$ clang-format -style=google test.cc
// Examples
// Regular multiline tokens
int x = f(R"(multi
line)",
2);
}
int y = g(h(R"(multi
line)"),
2);
// Formatted multiline tokens
int z = f(R"pb(multi: 1 #
line: 2)pb", 2);
int w = g(h(R"pb(multi: 1 #
line: 2)pb"), 2);
```
This patch addresses this inconsistency by forcing breaking after multiline formatted raw string literals. This requires a little tweak to the indentation chosen for the contents of a formatted raw string literal: in case when that's a parameter and not the last one, the indentation is based off of the uniform indentation of all of the parameters.
Richard Trieu [Thu, 25 Oct 2018 01:08:00 +0000 (01:08 +0000)]
[Sema] Fix -Wcomma for C89
There is a small difference in the scope flags for C89 versus the other C/C++
dialects. This change ensures that the -Wcomma warning won't be duplicated or
issued in the wrong location. Also, the test case is refactored into C and C++
parts, with the C++ parts guarded by a #ifdef to allow the test to run in both
modes.
Driver,CodeGen: introduce support for Swift CFString layout
Add a new driver level flag `-fcf-runtime-abi=` that allows one to specify the
runtime ABI for CoreFoundation. This controls the language interoperability.
In particular, this is relevant for generating the CFConstantString classes
(primarily through the `__builtin___CFStringMakeConstantString` builtin) which
construct a reference to the "CFObject"'s `isa` field. This type differs
between swift 4.1 and 4.2+.
Valid values for the new option include:
- objc [default behaviour] - enable ObjectiveC interoperability
- swift-4.1 - enable interoperability with swift 4.1
- swift-4.2 - enable interoperability with swift 4.2
- swift-5.0 - enable interoperability with swift 5.0
- swift [alias] - target the latest swift ABI
Furthermore, swift 4.2+ changed the layout for the CFString when building
CoreFoundation *without* ObjectiveC interoperability. In such a case, a field
was added to the CFObject base type changing it from: <{ const int*, int }> to
<{ uintptr_t, uintptr_t, uint64_t }>.
In swift 5.0, the CFString type will be further adjusted to change the length
from a uint32_t on everything but BE LP64 targets to uint64_t.
Note that the default behaviour for clang remains unchanged and the new layout
must be explicitly opted into via `-fcf-runtime-abi=swift*`.
Volodymyr Sapsai [Wed, 24 Oct 2018 22:39:38 +0000 (22:39 +0000)]
[VFS] Remove 'ignore-non-existent-contents' attribute for YAML-based VFS.
'ignore-non-existent-contents' stopped working after r342232 in a way
that the actual attribute value isn't used and it works as if it is
always `true`.
Common use case for VFS iteration is iterating through files in umbrella
directories for modules. Ability to detect if some VFS entries point to
non-existing files is nice but non-critical. Instead of adding back
support for `'ignore-non-existent-contents': false` I am removing the
attribute, because such scenario isn't used widely enough and stricter
checks don't provide enough value to justify the maintenance.
Eric Fiselier [Wed, 24 Oct 2018 22:38:49 +0000 (22:38 +0000)]
[SemaCXX] Unconfuse Clang when std::align_val_t is unscoped in C++03
Summary:
When -faligned-allocation is specified in C++03 libc++ defines std::align_val_t as an unscoped enumeration type (because Clang didn't provide scoped enumerations as an extension until 8.0).
Unfortunately Clang confuses the `align_val_t` overloads of delete with the sized deallocation overloads which aren't enabled. This caused Clang to call the aligned deallocation function as if it were the sized deallocation overload.
The first set of instructions corresponds to the first taskloop construct. It is important to note that the implicit taskgroup region associated with the taskloop construct has been materialized in our IR: the `__kmpc_taskloop` occurs inside a taskgroup region. Note also that this taskgroup region does not exist in our second taskloop because we are using the `nogroup` clause.
The issue here is the 4th argument of the kmpc_taskloop call, starting from the end, is always a zero. Checking the LLVM OpenMP RT implementation, we see that this argument corresponds to the nogroup parameter:
```
void __kmpc_taskloop(ident_t *loc, int gtid, kmp_task_t *task, int if_val,
kmp_uint64 *lb, kmp_uint64 *ub, kmp_int64 st, int nogroup,
int sched, kmp_uint64 grainsize, void *task_dup);
```
So basically we always tell to the RT to do another taskgroup region. For the first taskloop, this means that we create two taskgroup regions. For the second example, it means that despite the fact we had a nogroup clause we are going to have a taskgroup region, so we unnecessary wait until all descendant tasks have been executed.
Alexey Bataev [Wed, 24 Oct 2018 18:53:12 +0000 (18:53 +0000)]
[OPENMP]Fix PR39366: do not try to private field if it is not captured.
The compiler is crashing if we trying to post-capture the fields
implicitly captured inside of the task constructs. Seems, this kind of
processing is not supported and such fields should not be
firstprivatized.
[Hexagon] Flip hexagon-autohvx to be true by default
This will allow other generators of LLVM IR to use the auto-vectorizer
without having to change that flag.
Note: on its own, this patch will disable auto-vectorization on Hexagon
in all cases, regardless of the -fvectorize flag. There is a companion
LLVM patch that together with this one forms an NFC for clang users.
Craig Topper [Wed, 24 Oct 2018 17:42:17 +0000 (17:42 +0000)]
[CodeGen] Update min-legal-vector width based on function argument and return types
This is a continuation of my patches to inform the X86 backend about what the largest IR types are in the function so that we can restrict the backend type legalizer to prevent 512-bit vectors on SKX when -mprefer-vector-width=256 is specified if no explicit 512 bit vectors were specified by the user.
This patch updates the vector width based on the argument and return types of the current function and from the types of any functions it calls. This is intended to make sure the backend type legalizer doesn't disturb any types that are required for ABI.
Erich Keane [Wed, 24 Oct 2018 14:33:30 +0000 (14:33 +0000)]
Remove a pair of unused dispatch multiversion declarations.
These declarations somehow survived a cleanup that combined them with the target
multiversioning functions. This patch removes them as they are no
longer necessary or used.
Eric Liu [Wed, 24 Oct 2018 12:57:27 +0000 (12:57 +0000)]
[CodeComplete] Expose InBaseClass signal in code completion results.
Summary:
No new tests as the existing tests for result priority should give us
coverage. Also as the new flag is trivial enough, I'm reluctant to plumb the
flag to c-index-test output.
Yuka Takahashi [Wed, 24 Oct 2018 12:43:25 +0000 (12:43 +0000)]
[autocompletion] Handle the space before pressing tab
Summary:
Distinguish "--autocomplete=-someflag" and "--autocomplete=-someflag,"
because the latter indicates that the user put a space before pushing tab
which should end up in a file completion.
Yuka Takahashi [Wed, 24 Oct 2018 08:24:16 +0000 (08:24 +0000)]
[bash-autocompletion] Fix bug when a flag ends with '='
There was a bug that when a flag ends with '=' and no value was suggested,
clang autocompletes the flag itself.
For example, in bash, it looked like this:
```
$ clang -fmodule-file=[tab]
-> $clang -fmodule-file=-fmodule-file
```
This is not what we expect. We expect a file autocompletion when no value
was found. With this patch, pressing tab suggests files in the current
directory.
Richard Trieu [Wed, 24 Oct 2018 02:07:41 +0000 (02:07 +0000)]
[Sema] Fix -Wcomma in dependent context
When there is a dependent type inside a cast, the CastKind becomes CK_Dependent
instead of CK_ToVoid. This fix will check that there is a dependent cast,
the original type is dependent, and the target type is void to ignore the cast.
Adrian Prantl [Wed, 24 Oct 2018 00:06:02 +0000 (00:06 +0000)]
Debug Info (-gmodules): emit full types for non-anchored template specializations
Before this patch, clang would emit a (module-)forward declaration for
template instantiations that are not anchored by an explicit template
instantiation, but still are guaranteed to be available in an imported
module. Unfortunately detecting the owning module doesn't reliably
work when local submodule visibility is enabled and the template is
inside a cross-module namespace.
This make clang debuggable again with -gmodules and LSV enabled.
George Karpenkov [Tue, 23 Oct 2018 18:24:53 +0000 (18:24 +0000)]
[analyzer] Rename trackNullOrUndefValue to trackExpressionValue
trackNullOrUndefValue is a long and confusing name,
and it does not actually reflect what the function is doing.
Give a function a new name, with a relatively clear semantics.
Leonard Chan [Tue, 23 Oct 2018 17:55:35 +0000 (17:55 +0000)]
[Fixed Point Arithmetic] Fixed Point to Boolean Cast
This patch is a part of https://reviews.llvm.org/D48456 in an attempt to split
the casting logic up into smaller patches. This contains the code for casting
from fixed point types to boolean types.
Kadir Cetinkaya [Tue, 23 Oct 2018 13:49:37 +0000 (13:49 +0000)]
[clang] Fix a null pointer dereference.
Summary:
Sometimes expression inside switch statement can be invalid, for
example type might be incomplete. In those cases code were causing a null
pointer dereference. This patch fixes that.
Hans Wennborg [Tue, 23 Oct 2018 13:17:13 +0000 (13:17 +0000)]
Revert r345009 "[DebugInfo] Generate debug information for labels. (After fix PR39094)"
This broke the Chromium build. See
https://bugs.chromium.org/p/chromium/issues/detail?id=898152#c1 for the
reproducer.
> Generate DILabel metadata and call llvm.dbg.label after label
> statement to associate the metadata with the label.
>
> After fixing PR37395.
> After fixing problems in LiveDebugVariables.
> After fixing NULL symbol problems in AddressPool when enabling
> split-dwarf-file.
> After fixing PR39094.
>
> Differential Revision: https://reviews.llvm.org/D45045
Aleksandr Urakov [Tue, 23 Oct 2018 08:23:22 +0000 (08:23 +0000)]
[AST] Do not align virtual bases in `MicrosoftRecordLayoutBuilder` when
an external layout is used
Summary:
The patch removes alignment of virtual bases when an external layout is used.
We have two cases:
- the external layout source has an information about virtual bases offsets,
so we just use them;
- the external source has no information about virtual bases offsets. In this
case we can't predict where the base will be located. If we will align it but
there will be something like `#pragma pack(push, 1)` really, then likely our
layout will not fit into the real structure size, and then some asserts will
hit. The asserts look reasonable, so I don't think that we need to remove
them. May be it would be better instead don't align fields / bases etc.
(so treat it always as `#pragma pack(push, 1)`) when an external layout source
is used but no info about a field location is presented.
Hsiangkai Wang [Tue, 23 Oct 2018 08:06:21 +0000 (08:06 +0000)]
[DebugInfo] Generate debug information for labels. (After fix PR39094)
Generate DILabel metadata and call llvm.dbg.label after label
statement to associate the metadata with the label.
After fixing PR37395.
After fixing problems in LiveDebugVariables.
After fixing NULL symbol problems in AddressPool when enabling
split-dwarf-file.
After fixing PR39094.
Martin Storsjo [Tue, 23 Oct 2018 07:01:55 +0000 (07:01 +0000)]
Revert "[Driver] Use forward slashes in most linker arguments"
This reverts commit r345004, as it broke tests when actually run
on windows; see e.g.
http://lab.llvm.org:8011/builders/clang-x64-windows-msvc/builds/763.
This broke tests that had captured a variable containing a path
with backslashes, which failed to match cases in the output
where the path separators had been changed into forward slashes.
Martin Storsjo [Tue, 23 Oct 2018 06:33:26 +0000 (06:33 +0000)]
[Driver] Use forward slashes in most linker arguments
libtool inspects the output of $CC -v to detect what object files and
libraries are linked in by default. When clang is built as a native
windows executable, all paths are formatted with backslashes, and
the backslashes cause each argument to be enclosed in quotes. The
backslashes and quotes break further processing within libtool (which
is implemented in shell script, running in e.g. msys) pretty badly.
Between unix style pathes (that only work in tools that are linked
to the msys runtime, essentially the same as cygwin) and proper windows
style paths (with backslashes, that can easily break shell scripts
and msys environments), the best compromise is to use windows style
paths (starting with e.g. c:) but with forward slashes, which both
msys based tools, shell scripts and native windows executables can
cope with. This incidentally turns out to be the form of paths that
GCC prints out when run with -v on windows as well.
This change potentially makes the output from clang -v a bit more
inconsistent, but it is isn't necessarily very consistent to begin with.
Richard Trieu [Tue, 23 Oct 2018 01:26:28 +0000 (01:26 +0000)]
[CodeGen] Attach InlineHint to more functions
For instantiated functions, search the template pattern to see if it marked
inline to determine if InlineHint attribute should be added to the function.
Craig Topper [Tue, 23 Oct 2018 00:15:37 +0000 (00:15 +0000)]
[X86] Remove 'rtm' feature from KNL.
I'm unsure if KNL has this feature, but the backend never thought it did, only clang did. The predefined-arch-macros test lost the check for __RTM__ on KNL when it was removed Skylake CPUs in r344117.
I think we want to drop it from KNL for consistency with Skylake anyway regardless of how we got here.
Erich Keane [Mon, 22 Oct 2018 21:20:45 +0000 (21:20 +0000)]
Give Multiversion-inline functions linkonce linkage
Since multiversion variant functions can be inline, in C they become
available-externally linkage. This ends up causing the variants to not
be emitted, and not available to the linker.
The solution is to make sure that multiversion functions are always
emitted by marking them linkonce.
Nick Desaulniers [Mon, 22 Oct 2018 19:48:08 +0000 (19:48 +0000)]
[Driver] allow Android triples to alias for non Android targets
Summary:
Partial revert of r330873 ('[Driver] Reland "Android triples are not
aliases for other triples."')
While we don't want `-target *-linux-android` to alias to non
*-linux-android libs and binaries, it turns out we do want the
opposite. Ie. We would like for `-target *-linux-gnu` to still be
able to use *-android libs and binaries.
In fact, this is used to cross assemble and link the Linux kernel for
Android devices.
`-target *-linux-gnu` needs to be used for the Linux kernel when
using the android binutils prebuilts (*-linux-android).
The use of `-target *-linux-android` on C source files will cause
Clang to perform optimizations based on the presence of bionic (due to
r265481 ('Faster stack-protector for Android/AArch64.')) which is
invalid within the Linux kernel and will produce a non-bootable kernel
image.
Of course, you could just use the standard binutils (*-linux-gnu),
but Android does not distribute these. So this patch fixes a problem
that only occurs when cross assembling and linking a Linux kernel with
the Android provided binutils, which is what is done within Android's
build system.
Adrian Prantl [Mon, 22 Oct 2018 16:27:41 +0000 (16:27 +0000)]
Ensure sanitizer check function calls have a !dbg location
Function calls without a !dbg location inside a function that has a
DISubprogram make it impossible to construct inline information and
are rejected by the verifier. This patch ensures that sanitizer check
function calls have a !dbg location, by carrying forward the location
of the preceding instruction or by inserting an artificial location if
necessary.
This fixes a crash when compiling the attached testcase with -Os.
David Greene [Mon, 22 Oct 2018 13:46:12 +0000 (13:46 +0000)]
Always search sysroot for GCC installs
Previously, if clang was configured with -DGCC_INSTALL_PREFIX, then it
would not search a provided sysroot for a gcc install. This caused a
number of regression tests to fail. If a sysroot is given, skip
searching GCC_INSTALL_PREFIX as it is likely not valid for the
provided sysroot.
Peter Smith [Mon, 22 Oct 2018 10:40:52 +0000 (10:40 +0000)]
[ARM][AArch64] Add LLVM_FALLTHROUGH to silence warning [NFC]
A follow up to D52784 to add in LLVM_FALLTHROUGH where there is an
intentional fall through in a switch statement. This will hopefully silence
a GCC warning.
Kristof Umann [Sun, 21 Oct 2018 23:30:01 +0000 (23:30 +0000)]
[analyzer][UninitializedObjectChecker] No longer using nonloc::LazyCompoundVal
As rightly pointed out by @NoQ, nonloc::LazyCompoundVals were only used to acquire a constructed object's region, which isn't what LazyCompoundVal was made for.
Kristof Umann [Sun, 21 Oct 2018 18:19:32 +0000 (18:19 +0000)]
[analyzer][NFC] Fix inconsistencies in AnalyzerOptions
I'm in the process of refactoring AnalyzerOptions. The main motivation behind
here is to emit warnings if an invalid -analyzer-config option is given from the
command line, and be able to list them all.
This first NFC patch contains small modifications to make AnalyzerOptions.cpp a
little more consistent.
Aleksei Sidorin [Sat, 20 Oct 2018 22:49:23 +0000 (22:49 +0000)]
[AST, analyzer] Transform rvalue cast outputs to lvalues (fheinous-gnu-extensions)
Despite the fact that cast expressions return rvalues, GCC still
handles such outputs as lvalues when compiling inline assembler.
In this commit, we are treating it by removing LValueToRValue
casts inside GCCAsmStmt outputs.
Akira Hatanaka [Sat, 20 Oct 2018 05:45:01 +0000 (05:45 +0000)]
[CodeGen] Use the mangle context owned by CodeGenModule to correctly
mangle types of lambda objects captured by a block instead of creating a
new mangle context everytime a captured field type is mangled.
This fixes a bug in IRGen's block helper merging code that was
introduced in r339438 where two blocks capturing two distinct lambdas
would end up sharing helper functions and the block descriptor. This
happened because the ID number used to distinguish lambdas defined
in the same context is reset everytime a mangled context is created.
Richard Trieu [Sat, 20 Oct 2018 02:15:58 +0000 (02:15 +0000)]
Make -Wfor-loop-analysis work with C++17
For now, disable the "variable in loop condition not modified" warning to not
be emitted when there is a structured binding variable in the loop condition.
Craig Topper [Sat, 20 Oct 2018 01:30:00 +0000 (01:30 +0000)]
[X86] When checking the bits in cpu_features for function multiversioning dispatcher in the resolver, make sure all the required bits are set. Not just one of them
Summary:
The multiversioning code repurposed the code from __builtin_cpu_supports for checking if a single feature is enabled. That code essentially performed (_cpu_features & (1 << C)) != 0. But with the multiversioning path, the mask is no longer guaranteed to be a power of 2. So we return true anytime any one of the bits in the mask is set not just all of the bits.
The correct check is (_cpu_features & mask) == mask
Artem Dergachev [Sat, 20 Oct 2018 00:29:24 +0000 (00:29 +0000)]
[analyzer] Be more plugin-friendly by moving static locals into .cpp files.
The GDMIndex functions return a pointer that's used as a key for looking up
data, but addresses of local statics defined in header files aren't the same
across shared library boundaries and the result is that analyzer plugins
can't access this data.
Event types are uniqued by using the addresses of a local static defined
in a header files, but it isn't the same across shared library boundaries
and plugins can't currently handle ImplicitNullDerefEvents.
Richard Smith [Fri, 19 Oct 2018 19:01:34 +0000 (19:01 +0000)]
PR24164, PR39336: init-captures are not distinct full-expressions.
Rather, they are subexpressions of the enclosing lambda-expression, and
any temporaries in them are destroyed at the end of that
full-expression, or when the corresponding lambda-expression is
destroyed if they are lifetime-extended.
Hans Wennborg [Fri, 19 Oct 2018 16:19:52 +0000 (16:19 +0000)]
Java annotation declaration being handled correctly
Previously, Java annotation declarations (@interface AnnotationName) were being
handled as ObjC interfaces. This caused the brace formatting to mess up, so
that when you had a class with an interface defined in it, it would indent the
final brace of the class.
It used to format this class like so:
class A {
@interface B {}
}
But will now just skip the @interface and format it like so:
Balazs Keri [Fri, 19 Oct 2018 13:32:20 +0000 (13:32 +0000)]
[ASTImporter] Added error handling for AST import.
Summary:
The goal of this change is to make the ASTImporter::Import functions return
llvm::Expected instead of the imported type.
As first part the ASTNodeImporter visit functions are updated to return with
llvm::Expected. Various `import` functions are added to ASTNodeImporter to
simplify the code and have a common place for interface towards ASTImporter
(from ASTNodeImporter). There is some temporary code that is needed before
ASTImporter is updated.
Marco Antognini [Fri, 19 Oct 2018 09:01:37 +0000 (09:01 +0000)]
[OpenCL] Remove unwanted signedness conversion from tests
The get_kernel_* functions used in cl20-device-side-enqueue.cl all return
unsigned integers. This patch avoids undesired implicit conversions on the
returned values.
[COFF, ARM64] Enable unit test arm64-microsoft-status-reg.cpp only for aarch64 target
This should unbreak bots broken here:
http://lab.llvm.org:8011/builders/clang-cmake-x86_64-sde-avx512-linux/builds/14391
http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-ubuntu-fast/builds/38288
Aaron Ballman [Thu, 18 Oct 2018 17:42:41 +0000 (17:42 +0000)]
Add language standard aliases for -std=c18, -std=gnu18, and -std=iso9899:2018.
As described in D40225, the C17 standard was balloted and approved in 2017, but the ISO publication process delayed the actual publication until 2018. WG14 considers the release to be C17 and describes it as such, but users can still be confused by the publication year which is why -std=c18 adds value. These aliases map to c17 and are all supported by GCC 8.x with the same behavior. Note that the value of __STDC_VERSION__ remains at 201710L.
projects / clang / log