Richard Smith [Tue, 30 Aug 2016 19:13:18 +0000 (19:13 +0000)]
PR29166: when merging declarations with typedef names for linkage purposes,
don't assume that the anonymous struct will be part of the most recent
declaration of the typedef.
Richard Smith [Tue, 30 Aug 2016 19:06:26 +0000 (19:06 +0000)]
Unrevert r280035 now that the clang-cl bug it exposed has been fixed by
r280133. Original commit message:
C++ Modules TS: driver support for building modules.
This works as follows: we add --precompile to the existing gamut of options for
specifying how far to go when compiling an input (-E, -c, -S, etc.). This flag
specifies that an input is taken to the precompilation step and no further, and
this can be specified when building a .pcm from a module interface or when
building a .pch from a header file.
The .cppm extension (and some related extensions) are implicitly recognized as
C++ module interface files. If --precompile is /not/ specified, the file is
compiled (via a .pcm) to a .o file containing the code for the module (and then
potentially also assembled and linked, if -S, -c, etc. are not specified). We
do not yet suppress the emission of object code for other users of the module
interface, so for now this will only work if everything in the .cppm file has
vague linkage.
As with the existing support for module-map modules, prebuilt modules can be
provided as compiler inputs either via the -fmodule-file= command-line argument
or via files named ModuleName.pcm in one of the directories specified via
-fprebuilt-module-path=.
This also exposes the -fmodules-ts cc1 flag in the driver. This is still
experimental, and in particular, the concrete syntax is subject to change as
the Modules TS evolves in the C++ committee. Unlike -fmodules, this flag does
not enable support for implicitly loading module maps nor building modules via
the module cache, but those features can be turned on separately and used in
conjunction with the Modules TS support.
Olivier Goffart [Tue, 30 Aug 2016 17:42:29 +0000 (17:42 +0000)]
Fix colored diagnostics from tools
r271042 changed the way the diagnostic arguments are parsed. It assumes that
the diagnostics options were already parsed by the "Driver".
For tools using clang::Tooling, the diagnostics argument were not parsed.
Sjoerd Meijer [Tue, 30 Aug 2016 08:09:45 +0000 (08:09 +0000)]
This adds new options -fdenormal-fp-math and passes through option -ffast-math
to CC1, which are translated to function attributes and can e.g. be mapped on
build attributes FP_exceptions and FP_denormal. Setting these build attributes
allows better selection of floating point libraries.
Richard Smith [Tue, 30 Aug 2016 00:44:54 +0000 (00:44 +0000)]
C++ Modules TS: driver support for building modules.
This works as follows: we add --precompile to the existing gamut of options for
specifying how far to go when compiling an input (-E, -c, -S, etc.). This flag
specifies that an input is taken to the precompilation step and no further, and
this can be specified when building a .pcm from a module interface or when
building a .pch from a header file.
The .cppm extension (and some related extensions) are implicitly recognized as
C++ module interface files. If --precompile is /not/ specified, the file is
compiled (via a .pcm) to a .o file containing the code for the module (and then
potentially also assembled and linked, if -S, -c, etc. are not specified). We
do not yet suppress the emission of object code for other users of the module
interface, so for now this will only work if everything in the .cppm file has
vague linkage.
As with the existing support for module-map modules, prebuilt modules can be
provided as compiler inputs either via the -fmodule-file= command-line argument
or via files named ModuleName.pcm in one of the directories specified via
-fprebuilt-module-path=.
This also exposes the -fmodules-ts cc1 flag in the driver. This is still
experimental, and in particular, the concrete syntax is subject to change as
the Modules TS evolves in the C++ committee. Unlike -fmodules, this flag does
not enable support for implicitly loading module maps nor building modules via
the module cache, but those features can be turned on separately and used in
conjunction with the Modules TS support.
Igor Kudrin [Mon, 29 Aug 2016 11:48:50 +0000 (11:48 +0000)]
[Coverage] Prevent creating a redundant counter if a nested body ends with a macro.
If there were several nested statements arranged in a way that all of them
end up with the same macro, then the expansion of this macro was assigned
with all the corresponding counters of these statements.
As a result, the wrong counter value was shown for the macro in llvm-cov.
This patch fixes the issue by preventing adding a counter for an expanded
source range if it already has an assigned counter, which is expected
to come from the most specific statement.
Nico Weber [Fri, 26 Aug 2016 21:51:14 +0000 (21:51 +0000)]
clang-cl: Support MSVC2015's /validate-charset flag.
Clang always assumes that files are utf-8. If an invalidly encoded character is
used in an identifier, clang always errors. If it's used in a character
literal, clang warns Winvalid-source-encoding (on by default). Clang never
checks the encoding of things in comments (adding this seems like a nice
feature if it doesn't impact performance).
For cl.exe /utf-8 (which enables /validate-charset), if a bad character is used
in an identifier, it emits both an error and a warning. If it's used in a
literal or a comment, it emits a warning.
So mapping /validate-charset to -Winvalid-source-encoding seems like a fairly
decent fit.
Also makes -fexec-charset accept utf-8 case-insensitively.
Like https://reviews.llvm.org/D23807, but for execution-charset.
Also replace a few .lower() comparisons with equals_lower().
Manman Ren [Fri, 26 Aug 2016 17:16:46 +0000 (17:16 +0000)]
Don't diagnose non-modular includes when we are not compiling a module.
This is triggered when we are compiling an implementation of a module,
it has relative includes to a VFS-mapped module with umbrella headers.
Currently we will find the real path to headers under the umbrella directory,
but the umbrella directories are using virtual path.
Richard Smith [Fri, 26 Aug 2016 00:14:38 +0000 (00:14 +0000)]
C++ Modules TS: add frontend support for building pcm files from module
interface files. At the moment, all declarations (and no macros) are exported,
and 'export' declarations are not supported yet.
Richard Smith [Thu, 25 Aug 2016 18:26:30 +0000 (18:26 +0000)]
Refactor to remove the assumption that we know the name of the module we're emitting at the point when we create a PCHGenerator (with the C++ modules TS, we find that out part way through parsing the input).
Adrian McCarthy [Thu, 25 Aug 2016 18:24:35 +0000 (18:24 +0000)]
Omit column info for CodeView by default
Clang tracks only start columns, not start-end ranges. CodeView allows for that, but the VS debugger doesn't handle anything less than a complete range well--it either highlights the wrong part of a statement or truncates source lines in the assembly view. It's better to have no column information at all.
So by default, we'll omit the column information for CodeView targeting Windows.
Since the column info is still useful for sanitizers, I've promoted -gcolumn-info (and -gno-column-info) to a CoreOption and added a couple tests to make sure that works for clang-cl.
Reid Kleckner [Thu, 25 Aug 2016 18:23:28 +0000 (18:23 +0000)]
[MS] Pass non-trivially-copyable objects indirectly on Windows ARM
This isn't exactly what MSVC does, unfortunately. MSVC does not pass
objects with destructors but no copy constructors by address. More ARM
expertise is required to really understand what should be done here.
Richard Smith [Thu, 25 Aug 2016 00:34:00 +0000 (00:34 +0000)]
Lazily load the ContextDecl for a lambda's DefinitionData, to fix a
deserialization cycle caused by the ContextDecl recursively importing members
of the lambda's closure type.
David Blaikie [Wed, 24 Aug 2016 23:22:36 +0000 (23:22 +0000)]
DebugInfo: Let -gsplit-dwarf and -gmlt compose if -fno-split-dwarf-inlining is used
If the inline info is not duplicated into the skeleton CU, then there's
value in using -gsplit-dwarf and -gmlt together (to keep all those extra
subprograms out of the skeleton CU, while also producing smaller .dwo
files)
Richard Smith [Wed, 24 Aug 2016 21:30:00 +0000 (21:30 +0000)]
Disable test under asan: it uses a lot of stack, and asan increases the
per-frame stack usage enough to cause it to hit our stack limit. This is not
ideal; we should find a better way of dealing with this, such as increasing
our stack allocation when built with ASan.
[Sema][Comments] Support @param with c++ 'using' keyword
Give appropriate warnings with -Wdocumentation for @param comments
that refer to function aliases defined with 'using'. Very similar
to typedef's behavior. This does not add support for
TypeAliasTemplateDecl yet.
David Blaikie [Wed, 24 Aug 2016 18:29:58 +0000 (18:29 +0000)]
DebugInfo: Add flag to CU to disable emission of inline debug info into the skeleton CU
In cases where .dwo/.dwp files are guaranteed to be available, skipping
the extra online (in the .o file) inline info can save a substantial
amount of space - see the original r221306 for more details there.
Samuel Antao [Wed, 24 Aug 2016 15:39:07 +0000 (15:39 +0000)]
[Driver][OpenMP][CUDA] Add capability to bundle object files in sections of the host binary format.
Summary:
This patch adds the capability to bundle object files in sections of the host binary using a designated naming convention for these sections. This patch uses the functionality of the object reader already in the LLVM library to read bundled files, and invokes clang with the incremental linking options to create bundle files.
Bundling files involves creating an IR file with the contents of the bundle assigned as initializers of globals binded to the designated sections. This way the bundling implementation is agnostic of the host object format.
The features added by this patch were requested in the RFC discussion in http://lists.llvm.org/pipermail/cfe-dev/2016-February/047547.html.
Summary:
One of the goals of programming models that support offloading (e.g. OpenMP) is to enable users to offload with little effort, by annotating the code with a few pragmas. I'd also like to save users the trouble of changing their existent applications' build system. So having the compiler always return a single file instead of one for the host and each target even if the user is doing separate compilation is desirable.
This diff proposes a tool named clang-offload-bundler (happy to change the name if required) that is used to bundle files associated with the same user source file but different targets, or to unbundle a file into separate files associated with different targets.
This tool supports the driver support for OpenMP under review in http://reviews.llvm.org/D9888. The tool is used there to enable separate compilation, so that the very first action on input files that are not source files is a "unbundling action" and the very last non-linking action is a "bundling action".
The format of the bundled files is currently very simple: text formats are concatenated with comments that have a magic string and target identifying triple in between, and binary formats have a header that contains the triple and the offset and size of the code for host and each target.
The goal is to improve this tool in the future to deal with archive files so that each individual file in the archive is properly dealt with. We see that archives are very commonly used in current applications to combine separate compilation results. So I'm convinced users would enjoy this feature.
Chris Bieneman [Tue, 23 Aug 2016 20:07:07 +0000 (20:07 +0000)]
driver: Support checking for rlimits via cmake (when bootstrapping)
Summary:
Add a cmake check for sys/resource.h and replace the __has_include() check with its result, in order to make it possible to use rlimits when building with compilers not supporting __has_include() -- i.e. when bootstrapping.
Richard Smith [Tue, 23 Aug 2016 19:41:39 +0000 (19:41 +0000)]
Fix regression introduced by r279164: only pass definitions as the PatternDef
to DiagnoseUninstantiableTemplate, teach hasVisibleDefinition to correctly
determine whether a function definition is visible, and mark both the function
and the template as visible when merging function template definitions to
provide hasVisibleDefinition with the relevant information.
The change to always pass the right declaration as the PatternDef to
DiagnoseUninstantiableTemplate also caused those checks to happen before other
diagnostics in InstantiateFunctionDefinition, giving worse diagnostics for the
same situations, so I sunk the relevant diagnostics into
DiagnoseUninstantiableTemplate. Those parts of this patch are based on changes
in reviews.llvm.org/D23492 by Vassil Vassilev.
This reinstates r279486, reverted in r279500, with a fix to
DiagnoseUninstantiableTemplate to only mark uninstantiable explicit
instantiation declarations as invalid if we actually diagnosed them. (When we
trigger an explicit instantiation of a class member from an explicit
instantiation declaration for the class, it's OK if there is no corresponding
definition and we certainly don't want to mark the member invalid in that
case.) This previously caused a build failure during bootstrap.
Tim Northover [Tue, 23 Aug 2016 18:12:58 +0000 (18:12 +0000)]
ARM-Darwin: ignore and diagnose attempts to omit frame pointer.
iOS (and other 32-bit ARM variants) always require a valid frame pointer to
improve backtraces. Previously the -fomit-frame-pointer and
-momit-leaf-frame-pointer options were being silently discarded via hacks in
the backend. It's better if Clang configures itself to emit the correct IR and
warns about (ignored) attempts to override this.
clang already treats all inputs as utf-8. Warn if anything but utf-8 is passed.
Do this by mapping source-charset to finput-charset, which already behaves like
this. Slightly tweak finput-charset to accept "utf-8" case-insensitively. This
matches gcc's and cl.exe's behavior, and IANA says that character set names are
case-insensitive.
Artem Dergachev [Tue, 23 Aug 2016 16:42:00 +0000 (16:42 +0000)]
[analyzer] Fix CloneDetector crash on calling methods of class templates.
If a call expression represents a method call of a class template,
and the method itself isn't templated, then the method may be considered
to be a template instantiation without template specialization arguments.
No longer crash when we could not find template specialization arguments.
Richard Smith [Mon, 22 Aug 2016 22:25:03 +0000 (22:25 +0000)]
Fix regression introduced by r279164: only pass definitions as the PatternDef
to DiagnoseUninstantiableTemplate, teach hasVisibleDefinition to correctly
determine whether a function definition is visible, and mark both the function
and the template as visible when merging function template definitions to
provide hasVisibleDefinition with the relevant information.
The change to always pass the right declaration as the PatternDef to
DiagnoseUninstantiableTemplate also caused those checks to happen before other
diagnostics in InstantiateFunctionDefinition, giving worse diagnostics for the
same situations, so I sunk the relevant diagnostics into
DiagnoseUninstantiableTemplate. Those parts of this patch are based on changes
in reviews.llvm.org/D23492 by Vassil Vassilev.
Adrian Prantl [Mon, 22 Aug 2016 22:23:58 +0000 (22:23 +0000)]
Module debug info: Don't assert when encountering an incomplete definition
in isDefinedInClangModule() and assume that the incomplete definition
is not defined in the module.
This broke the -gmodules self host recently.
rdar://problem/27894367
[SemaObjC] Do not RebuildObjCMessageExpr without valid method decl
Fix crash-on-invalid in ObjC Sema by avoiding to rebuild a message
expression to a 'super' class in case the method to call does not exist
(i.e. comes from another missing identifier).
In this case, the typo transform is invoked upon the message expression
in an attempt to solve a typo in a 'super' call parameters, but it
crashes since it assumes the method to call has a valid declaration.
Matt Arsenault [Mon, 22 Aug 2016 19:25:59 +0000 (19:25 +0000)]
AMDGPU: Handle structs directly in AMDGPUABIInfo
Structs are currently handled as pointer + byval, which makes AMDGPU
LLVM backend generate incorrect code when structs are used. This patch
changes struct argument to be handled directly and without flattening,
which Clover (Mesa 3D Gallium OpenCL state tracker) will be able to
handle. Flattening would expand the struct to individual elements and
pass each as a separate argument, which Clover can not
handle. Furthermore, such expansion does not fit the OpenCL
programming model which requires to explicitely specify each argument
index, size and memory location.
Artem Dergachev [Sat, 20 Aug 2016 17:35:53 +0000 (17:35 +0000)]
[analyzer] Use faster hashing (MD5) in CloneDetector.
This replaces the old approach of fingerprinting every AST node into a string,
which avoided collisions and was simple to implement, but turned out to be
extremely ineffective with respect to both performance and memory.
The collisions are now dealt with in a separate pass, which no longer causes
performance problems because collisions are rare.
Benjamin Kramer [Sat, 20 Aug 2016 16:51:33 +0000 (16:51 +0000)]
[Sema] Don't crash on scanf on forward-declared enums.
This is valid in GNU C, which allows pointers to incomplete enums. GCC
just pretends that the underlying type is 'int' in those cases, follow
that behavior.
Artem Dergachev [Sat, 20 Aug 2016 10:06:59 +0000 (10:06 +0000)]
[analyzer] Make CloneDetector consider macro expansions.
So far macro-generated code was treated by the CloneDetector as normal code.
This caused that some macros where reported as false-positive clones because
large chunks of code coming from otherwise concise macro expansions were treated
as copy-pasted code.
This patch ensures that macros are treated in the same way as literals/function
calls. This prevents macros that expand into multiple statements
from being reported as clones.
Manman Ren [Sat, 20 Aug 2016 00:04:21 +0000 (00:04 +0000)]
[NFC] Add a class ObjCProtocolQualifiers to wrap APIs for ObjC protocol list.
This is in preparation of adding a new type class ObjCTypeParamType that
can take protocol qualifiers. ObjCProtocolQualifiers will be shared between
ObjCObjectType and ObjCTypeParamType.