Rafael Espindola [Fri, 23 Feb 2018 19:30:48 +0000 (19:30 +0000)]
Bring r325915 back.
The tests that failed on a windows host have been fixed.
Original message:
Start setting dso_local for COFF.
With this there are still some GVs where we don't set dso_local
because setGVProperties is never called. I intend to fix that in
followup commits. This is just the bare minimum to teach
shouldAssumeDSOLocal what it should do for COFF.
Rafael Espindola [Fri, 23 Feb 2018 15:32:32 +0000 (15:32 +0000)]
Start setting dso_local for COFF.
With this there are still some GVs where we don't set dso_local
because setGVProperties is never called. I intend to fix that in
followup commits. This is just the bare minimum to teach
shouldAssumeDSOLocal what it should do for COFF.
Rafael Espindola [Fri, 23 Feb 2018 00:22:15 +0000 (00:22 +0000)]
Simplify setting dso_local. NFC.
The value of dso_local can be computed from just IR properties and
global information (object file type, command line options, etc).
With this patch we no longer pass in the Decl. It was almost unused
and making it fully unused guarantees that dso_local is consistent
with the rest of the IR.
This is a bug fix that removes the emission of reduction support for pragma 'distribute' when found alone or in combinations without simd.
Pragma 'distribute' does not have a reduction clause, but when combined with pragma 'simd' we need to emit the support for simd's reduction clause as part of code generation for distribute. This guard is similar to the one used for reduction support earlier in the same code gen function.
Jonas Hahnfeld [Thu, 22 Feb 2018 17:06:35 +0000 (17:06 +0000)]
[docs] Improve help for OpenMP options, NFC.
* Add HelpText for -fopenmp so that it appears in clang --help.
* Hide -fno-openmp-simd, only list the positive option.
* Hide -fopenmp-relocatable-target and -fopenmp-use-tls from
clang --help and from ClangCommandLineReference.
* Improve MetaVarName for -Xopenmp-target=<...>.
Summary:
OpenCL 2.0 specification defines '-cl-uniform-work-group-size' option,
which requires that the global work-size be a multiple of the work-group
size specified to clEnqueueNDRangeKernel and allows optimizations that
are made possible by this restriction.
The patch introduces the support of this option.
To keep information about whether an OpenCL kernel has uniform work
group size or not, clang generates 'uniform-work-group-size' function
attribute for every kernel:
- "uniform-work-group-size"="true" for OpenCL 1.2 and lower,
- "uniform-work-group-size"="true" for OpenCL 2.0 and higher if
'-cl-uniform-work-group-size' option was specified,
- "uniform-work-group-size"="false" for OpenCL 2.0 and higher if no
'-cl-uniform-work-group-size' options was specified.
If the function is not an OpenCL kernel, 'uniform-work-group-size'
attribute isn't generated.
Richard Trieu [Thu, 22 Feb 2018 05:32:25 +0000 (05:32 +0000)]
[ODRHash] Handle some template weirdness.
Build the index off of DeclarationName instead of Decl pointers. When finding
an UnresolvedLookupExprClass, hash it as if it were a DeclRefExpr. This will
allow methods to be hashed.
When using blocks with C++ on Windows x86, it is possible to have the
block literal be pushed into the inalloca'ed parameters. Teach IRGen to
handle the case properly by extracting the block literal from the
inalloca parameter. This fixes the use of blocks with C++ on Windows
x86.
Ben Hamilton [Wed, 21 Feb 2018 21:27:27 +0000 (21:27 +0000)]
[clang-format] Fix regression when getStyle() called with empty filename
Summary:
D43522 caused an assertion failure when getStyle() was called with
an empty filename:
P8065
This adds a test to reproduce the failure and fixes the issue by
ensuring we never pass an empty filename to
Environment::CreateVirtualEnvironment().
Test Plan: New test added. Ran test with:
% make -j12 FormatTests && ./tools/clang/unittests/Format/FormatTests
Before diff, test failed with P8065. Now, test passes.
Peter Szecsi [Wed, 21 Feb 2018 16:06:56 +0000 (16:06 +0000)]
[analyzer] Prevent AnalyzerStatsChecker from crash
The checker marks the locations where the analyzer creates sinks. However, it
can happen that the sink was created because of a loop which does not contain
condition statement, only breaks in the body. The exhausted block is the block
which should contain the condition but empty, in this case.
This change only emits this marking in order to avoid the undefined behavior.
Mikhail Maltsev [Wed, 21 Feb 2018 10:08:18 +0000 (10:08 +0000)]
[Sema] Classify conversions from enum to float as narrowing
Summary:
According to [dcl.init.list]p7:
A narrowing conversion is an implicit conversion
- ...
- from an integer type or unscoped enumeration type to a
floating-point type, except where the source is a constant
expression and the actual value after conversion will fit into
the target type and will produce the original value when
converted back to the original type, or
- ...
Currently clang does not handle the 'unscoped enumeration' case. This
patch fixes the corresponding check.
Serge Pavlov [Wed, 21 Feb 2018 02:02:39 +0000 (02:02 +0000)]
Clean up use of C allocation functions
If the value returned by `malloc`, `calloc` or `realloc` is not checked
for null pointer, this change replaces them for `safe_malloc`,
`safe_calloc` or `safe_realloc`, which are defined in the namespace `llvm`.
These function report fatal error on out of memory.
In the plain C files, assertion statements are added to ensure that memory
is successfully allocated.
The aim of this change is to get better diagnostics of OOM on Windows.
Simon Dardis [Wed, 21 Feb 2018 00:05:05 +0000 (00:05 +0000)]
[mips] Spectre variant two mitigation for MIPSR2
This patch provides mitigation for CVE-2017-5715, Spectre variant two,
which affects the P5600 and P6600. It provides the option
-mindirect-jump=hazard, which instructs the LLVM backend to replace
indirect branches with their hazard barrier variants.
This option is accepted when targeting MIPS revision two or later.
The migitation strategy suggested by MIPS for these processors is to
use two hazard barrier instructions. 'jalr.hb' and 'jr.hb' are hazard
barrier variants of the 'jalr' and 'jr' instructions respectively.
These instructions impede the execution of instruction stream until
architecturally defined hazards (changes to the instruction stream,
privileged registers which may affect execution) are cleared. These
instructions in MIPS' designs are not speculated past.
These instructions are used with the option -mindirect-jump=hazard
when branching indirectly and for indirect function calls.
These instructions are defined by the MIPS32R2 ISA, so this mitigation
method is not compatible with processors which implement an earlier
revision of the MIPS ISA.
Implementation note: I've opted to provide this as an
-mindirect-jump={hazard,...} style option in case alternative
mitigation methods are required for other implementations of the MIPS
ISA in future, e.g. retpoline style solutions.
Richard Smith [Tue, 20 Feb 2018 23:17:41 +0000 (23:17 +0000)]
When multiple sanitizers are enabled (ubsan + something else), use all relevant blacklists.
Ideally, we'd only use the ubsan blacklist for ubsan sanitizers, and only use
the other-sanitizer blacklist for its sanitizers, but this at least enables the
intended suppressions.
Erich Keane [Tue, 20 Feb 2018 22:25:28 +0000 (22:25 +0000)]
[NFC] In Multiversion Check function, switch to return Diag
This function did a lot of 'Diag, return true' stuff. This resulted in needing
to introduce scopes in quite a few places. This patch replaces useages of a
single "S.Diag" followed by return true with simply "return S.Diag".
Kamil Rytarowski [Tue, 20 Feb 2018 16:27:28 +0000 (16:27 +0000)]
Stop linking sanitized applications with -lutil and -lkvm on NetBSD
The proper approach is to rebuild libutil and libkvm with a desired sanitizer.
An alternative approach to reimplement these functions (and other ones like
curses(3), editline(3) etc) does not scale and enforces linkage every single
binary with these libraries.
Hans Wennborg [Tue, 20 Feb 2018 12:43:02 +0000 (12:43 +0000)]
Revert r325375 "[MS] Make constexpr static data members implicitly inline"
This broke Clang bootstrap on Windows, PR36453.
> This handles them exactly the same way that we handle const integral
> static data members with inline definitions, which is what MSVC does.
>
> As a follow-up, now that we have a way to mark variables inline in the
> AST, we should consider marking them implicitly inline there instead of
> only treating them as inline in CodeGen. Unfortunately, this breaks a
> lot of dllimport test cases, so that is future work for now.
>
> Fixes PR36125.
Ivan A. Kosarev [Tue, 20 Feb 2018 12:33:04 +0000 (12:33 +0000)]
[CodeGen] Fix generation of TBAA tags for may-alias accesses
This patch fixes creating TBAA access descriptors for
may_alias-marked access types. Currently, for such types we
generate ordinary descriptors with char as its access type. The
patch changes this to produce proper may-alias descriptors.
Richard Smith [Tue, 20 Feb 2018 02:32:30 +0000 (02:32 +0000)]
Fix some -Wexceptions false positives.
Reimplement the "noexcept function actually throws" warning to properly handle
nested try-blocks. In passing, change 'throw;' handling to treat any enclosing
try block as being sufficient to suppress the warning rather than requiring a
'catch (...)'; the warning is intended to be conservatively-correct.
[clang-format] Fix text proto extension scope opening detection
Summary:
This fixes the detection of scope openers in text proto extensions; previously
they were not detected correctly leading to instances like:
```
msg {
[aa.bb
] {
key: value
}
}
```
Ilya Biryukov [Mon, 19 Feb 2018 13:53:49 +0000 (13:53 +0000)]
[CodeComplete] Avoid name clashes of 'Kind' inside CodeCompletionContext. NFC
CodeCompletionContext had declarations of field and enum inside, both named 'Kind'.
It caused gcc 4.8 to give an incorrent warning when refering to enum as
`enum CodeCompletionContext::Kind`.
Avoid that warning by renaming the private field to CCKind.
Keith Walker [Mon, 19 Feb 2018 12:40:26 +0000 (12:40 +0000)]
[ARM] disable FPU features when using soft floating point.
To be compatible with GCC if soft floating point is in effect any FPU
specified is effectively ignored, eg,
-mfloat-abi=soft -fpu=neon
If any floating point features which require FPU hardware are enabled
they must be disable.
There was some support for doing this for NEON, but it did not handle
VFP, nor did it prevent the backend from emitting the build attribute
Tag_FP_arch describing the generated code as using the floating point
hardware if a FPU was specified (even though soft float does not use
the FPU).
Disabling the hardware floating point features for targets which are
compiling for soft float has meant that some tests which were incorrectly
checking for hardware support also needed to be updated. In such cases,
where appropriate the tests have been updated to check compiling for
soft float and a non-soft float variant (usually softfp). This was
usually because the target specified in the test defaulted to soft float.
Ivan A. Kosarev [Mon, 19 Feb 2018 09:49:11 +0000 (09:49 +0000)]
[CodeGen] Initialize large arrays by copying from a global
Currently, clang compiles explicit initializers for array
elements into series of store instructions. For large arrays of
built-in types this results in bloated output code and
significant amount of time spent on the instruction selection
phase. This patch fixes the issue by initializing such arrays
with global constants that store the binary image of the
initializer.
Dimitry Andric [Sat, 17 Feb 2018 21:04:35 +0000 (21:04 +0000)]
[X86] Add 'sahf' CPU feature to frontend
Summary:
Make clang accept `-msahf` (and `-mno-sahf`) flags to activate the
`+sahf` feature for the backend, for bug 36028 (Incorrect use of
pushf/popf enables/disables interrupts on amd64 kernels). This was
originally submitted in bug 36037 by Jonathan Looney
<jonlooney@gmail.com>.
As described there, GCC also uses `-msahf` for this feature, and the
backend already recognizes the `+sahf` feature. All that is needed is to
teach clang to pass this on to the backend.
The mapping of feature support onto CPUs may not be complete; rather, it
was chosen to match LLVM's idea of which CPUs support this feature (see
lib/Target/X86/X86.td).
I also updated the affected test case (CodeGen/attr-target-x86.c) to
match the emitted output.
Vitaly Buka [Fri, 16 Feb 2018 23:38:22 +0000 (23:38 +0000)]
[ThinLTO] Allow indexing to request backend to ignore the module
Summary:
Gold plugin does not add pass to ThinLTO modules without useful symbols.
In this case ThinLTO can't create corresponding index file and some features, like CFI,
cannot be processes by backed correctly without index.
Given that we don't need the backed output we can request it to avoid
processing the module. This is implemented by this patch using new
"SkipModuleByDistributedBackend" flag.
Vitaly Buka [Fri, 16 Feb 2018 23:34:16 +0000 (23:34 +0000)]
[ThinLTO] Ignore object files with no ThinLTO modules if -fthinlto-index= is set
Summary:
ThinLTO compilation may decide not to split module and keep at as regular LTO.
In this can this module already processed during indexing and already a part of
merged object file. So here we can just skip it.
Reid Kleckner [Fri, 16 Feb 2018 19:44:47 +0000 (19:44 +0000)]
[MS] Make constexpr static data members implicitly inline
This handles them exactly the same way that we handle const integral
static data members with inline definitions, which is what MSVC does.
As a follow-up, now that we have a way to mark variables inline in the
AST, we should consider marking them implicitly inline there instead of
only treating them as inline in CodeGen. Unfortunately, this breaks a
lot of dllimport test cases, so that is future work for now.
Erich Keane [Fri, 16 Feb 2018 17:31:59 +0000 (17:31 +0000)]
Clean up 'target' attribute diagnostics
There were a few issues previously with the target
attribute diagnostics implementation that lead to the
attribute being added to the AST despite having an error
in it.
This patch changes that, and adds a test to ensure it
does not get added to the AST.
[DOXYGEN] There was a request in the review D41507 to change the notation for hex numbers in doxygen documentation from <...>h to 0x<...>. Both of these notations were used in x86 intrinsics documentation. I promised to change them to 0x<...> for consistency.
[Modules] Extend -fmodule-name semantic for frameworks with private modules
Assume Foo.framework with two module maps and two modules Foo and
Foo_Private.
Framework authors need to skip building both Foo and Foo_Private when
using -fmodule-name=Foo, since both are part of the framework and used
interchangeably during compilation.
Alexey Bataev [Thu, 15 Feb 2018 23:39:43 +0000 (23:39 +0000)]
[OPENMP] Fix PR36399: Crash on C code with ordered doacross construct.
Codegen for ordered with doacross construct might produce incorrect code
because of missing cleanup scope for the construct. Without this scope
the final runtime function call could be emitted in the wrong order that
leads to incorrect codegen.
Brian Gesiak [Thu, 15 Feb 2018 20:37:22 +0000 (20:37 +0000)]
[Coroutines] Use allocator overload when available
Summary:
Depends on https://reviews.llvm.org/D42605.
An implementation of the behavior described in `[dcl.fct.def.coroutine]/7`:
when a promise type overloads `operator new` using a "placement new"
that takes the same argument types as the coroutine function, that
overload is used when allocating the coroutine frame.
Simply passing references to the coroutine function parameters directly
to `operator new` results in invariant violations in LLVM's coroutine
splitting pass, so this implementation modifies Clang codegen to
produce allocator-specific alloc/store/loads for each parameter being
forwarded to the allocator.
Brian Gesiak [Thu, 15 Feb 2018 20:09:25 +0000 (20:09 +0000)]
[Sema] Toggle diags when finding allocators (NFCI)
Summary:
Many methods in Sema take a `bool Diagnose` parameter. Examples of such
methods include `Sema::FindDeallocationFunction` and
`Sema::SpecialMemberIsTrivial`. Calling these methods with
`Diagnose = false` allows callers to, for instance, check for the
existence of a deallocation function, without that check resulting in
error diagnostics being emitted if no matching deallocation function exists.
Add a similar `bool Diagnose` to the `Sema::FindAllocationFunctions`
method, so that checks for the existence of allocation functions can be
made without triggering error diagnostics.
This allows `SemaCoroutine.cpp`, in its implementation of the
Coroutines TS, to check for the existence of a particular `operator new`
overload, but then without error fall back to a default `operator new`
if no matching overload exists.
Artem Dergachev [Thu, 15 Feb 2018 19:34:19 +0000 (19:34 +0000)]
[analyzer] Suppress temporary destructors for temporary arrays.
Array destructors, like constructors, need to be called for each element of the
array separately. We do not have any mechanisms to do this in the analyzer,
so for now all we do is evaluate a single constructor or destructor
conservatively and give up. It automatically causes the necessary invalidation
and pointer escape for the whole array, because this is how RegionStore works.
Implement this conservative behavior for temporary destructors. This fixes the
crash on the provided test.
Artem Dergachev [Thu, 15 Feb 2018 19:28:21 +0000 (19:28 +0000)]
[analyzer] Implement path notes for temporary destructors.
Temporary destructors fire at the end of the full-expression. It is reasonable
to attach the path note for entering/leaving the temporary destructor to its
CXXBindTemporaryExpr. This would not affect lifetime-extended temporaries with
their automatic destructors which aren't temporary destructors.
The path note may be confusing in the case of destructors after elidable copy
constructors.
Artem Dergachev [Thu, 15 Feb 2018 19:17:44 +0000 (19:17 +0000)]
[analyzer] Compute the correct this-region for temporary destructors.
Inline them if possible - a separate flag is added to control this.
The whole thing is under the cfg-temporary-dtors flag, off by default so far.
Temporary destructors are called at the end of full-expression. If the
temporary is lifetime-extended, automatic destructors kick in instead,
which are not addressed in this patch, and normally already work well
modulo the overally broken support for lifetime extension.
The patch operates by attaching the this-region to the CXXBindTemporaryExpr in
the program state, and then recalling it during destruction that was triggered
by that CXXBindTemporaryExpr. It has become possible because
CXXBindTemporaryExpr is part of the construction context since r325210.
projects / clang / log