Manuel Klimek [Tue, 25 Jul 2017 10:22:06 +0000 (10:22 +0000)]
Fix incorrect use of current directory to find moved paths in ASTReader.
CurrentDir was set as the path of the current module, but that can change as
part of a chain of loaded modules.
When we try to locate a file mentioned in a module that does not exist, we use
a heuristic to look at the relative path between the original location of the
module and the file we look for, and use that relatively to the CurrentDir.
This only works if CurrentDir is the same as the (current) path of the module
file the file was mentioned in; if it is not, we look at the path relatively to
the wrong directory, and can end up reading random unrelated files that happen
to have the same name.
This patch fixes this by using the BaseDirectory of the module file the file
we look for was mentioned in instead of the CurrentDir heuristic.
[analyzer] Treat throws as sinks for suppress-on-sink purposes.
Because since r308957 the suppress-on-sink feature contains its own
mini-analysis, it also needs to become aware that C++ unhandled exceptions
cause sinks. Unfortunately, for now we treat all exceptions as unhandled in
the analyzer, so suppress-on-sink needs to do the same.
[analyzer] Further improve suppress-on-sink behavior in incomplete analyses.
If a certain memory leak (or other similar bug) found by the analyzer is known
to be happening only before abnormal termination of the program ("sink", eg.
assertion failure in the code under analysis, or another bug that introduces
undefined behavior), such leak warning is discarded. However, if the analysis
has never reaches completion (due to complexity of the code), it may be
failing to notice the sink.
This commit further extends the partial solution introduced in r290341 to cover
cases when a complicated control flow occurs before encountering a no-return
statement (which anyway inevitably leads to such statement(s)) by traversing
the respective section of the CFG in a depth-first manner. A complete solution
still seems elusive.
[NFC] Use RAII to un-poison and then re-poison __VA_ARGS__
- This will also be used for the forthcoming __VA_OPT__ feature approved for C++2a.
- recommended by rsmith during his review of the __VA_OPT__ patch (https://reviews.llvm.org/D35782)
Adrian Prantl [Mon, 24 Jul 2017 23:48:51 +0000 (23:48 +0000)]
Debug Info: Avoid completing class types when a definition is in a module.
This patch adds an early exit to CGDebugInfo::completeClassData() when
compiling with -gmodules and the to-be-completed type is available in
a clang module.
Under Windows Itanium, we need to export virtual and non-virtual thunks
if the functions being thunked are exported. These thunks would
previously inherit their dllexport attribute from the declaration, but
r298330 changed declarations to not have dllexport attributes. We
therefore need to add the dllexport attribute to the definition
ourselves now. This is consistent with MinGW GCC's behavior.
This redoes r306770 but limits the logic to Itanium. MicrosoftCXXABI's
setThunkLinkage ensures that thunks aren't exported under that ABI, so
I'm handling this in ItaniumCXXABI's setThunkLinkage for symmetry.
We need to export these thunks because they can be referenced outside
the library they're defined in. For example, if a child class without a
key function inherits from a parent class with a key function, the
parent's thunks will only be defined in the library with the key
function, but the construction vtable for the parent in the child might
be emitted outside the library (since the child doesn't have a key
function), and it needs to reference the parent's thunks.
We don't need to mark these thunks as imported since any references to
them will occur in data, so the compiler can't generate the IAT load
sequence anyway. Instead, we rely on the linker generating import thunks
for the thunks.
C2017 update 3 produces a clang that crashes when compiling clang. Disabling
optimizations for StmtProfiler::VisitCXXOperatorCallExpr() makes the crash go
away.
Patch from Bruce Dawson <brucedawson@chromium.org>!
https://reviews.llvm.org/D35757
[clang-format] Fix comment levels between '} else {' and PPDirective.
Summary:
This fixes a regression exposed by r307795 and rL308725 in which the level of a
comment line between '} else {' and a preprocessor directive is incorrectly set
as the level of the '} else {' line. For example, this :
```
int f(int i) {
if (i) {
++i;
} else {
// comment
#ifdef A
--i;
#endif
}
}
```
was formatted as:
```
int f(int i) {
if (i) {
++i;
} else {
// comment
#ifdef A
--i;
#endif
}
}
```
[Modules] Rework r274270. Let Clang targets depend on intrinsics_gen.
This gets rid of almost LLVM targets unconditionally depending on intrinsic_gen.
Clang's modules still have weird dependencies and hard to remove intrinsics_gen in better way.
Then, it'd be better to give whole clang targets depend on intrinsic_gen.
Erich Keane [Fri, 21 Jul 2017 22:37:03 +0000 (22:37 +0000)]
Break up Targets.cpp into a header/impl pair per target type[NFCI]
Targets.cpp is getting unwieldy, and even minor changes cause the entire thing
to cause recompilation for everyone. This patch bites the bullet and breaks
it up into a number of files.
I tended to keep function definitions in the class declaration unless it
caused additional includes to be necessary. In those cases, I pulled it
over into the .cpp file. Content is copy/paste for the most part,
besides includes/format/etc.
Erich Keane [Fri, 21 Jul 2017 18:50:36 +0000 (18:50 +0000)]
Complex Long Double classification In RegCall calling convention
This change is part of the RegCall calling convention support for LLVM.
Existing RegCall implementation was extended to include correct handling of
Complex Long Double type. Complex long double types should be returned/passed
in memory and not register stack. This patch implements this behavior.
Alex Lorenz [Fri, 21 Jul 2017 12:49:28 +0000 (12:49 +0000)]
[clang-diff] Add initial implementation
This is the first commit for the "Clang-based C/C++ diff tool" GSoC project.
ASTDiff is a new library that computes a structural AST diff between two ASTs
using the gumtree algorithm. Clang-diff is a new Clang tool that will show
the structural code changes between different ASTs.
[clang-format] Fix comment levels between '}' and PPDirective
Summary:
This fixes a regression exposed by r307795 in which the level of a comment line
between '}' and a preprocessor directive is incorrectly set as the level of the
line before the '}'. In effect, this:
```
int f(int i) {
int j = i;
return i + j;
}
// comment
#ifdef A
#endif
```
was formatted as:
```
int f(int i) {
int j = i;
return i + j;
}
// comment
Simon Atanasyan [Fri, 21 Jul 2017 08:10:57 +0000 (08:10 +0000)]
[mips] Add `short_call` to the set of `long_call/far/near` attributes
MIPS gcc supports `long_call/far/near` attributes only, but other
targets have the `short_call` attribut, so let's support it for MIPS
for consistency.
Richard Trieu [Fri, 21 Jul 2017 04:56:48 +0000 (04:56 +0000)]
Fix tblgen error.
tblgen couldn't determing a unique name between "long_call" and "far", so it
errored out when generating documentation. Copy the documentation, and give
an explicit header for "long_call".
Petr Hosek [Fri, 21 Jul 2017 01:17:49 +0000 (01:17 +0000)]
[Driver] Consider -fno-sanitize=... state when filtering out -fsanitize-coverage=...
The driver ignores -fsanitize-coverage=... flags when also given
-fsanitize=... flags for sanitizer flavors that don't support the
coverage runtime. This logic failed to account for subsequent
-fno-sanitize=... flags that disable the sanitizer flavors that
conflict with -fsanitize-coverage=... flags.
The patch adds support of i128 params lowering. The changes are quite trivial to
support i128 as a "special case" of integer type. With this patch, we lower i128
params the same way as aggregates of size 16 bytes: .param .b8 _ [16].
Currently, NVPTX can't deal with the 128 bit integers:
* in some cases because of failed assertions like
ValVTs.size() == OutVals.size() && "Bad return value decomposition"
* in other cases emitting PTX with .i128 or .u128 types (which are not valid [1])
[1] http://docs.nvidia.com/cuda/parallel-thread-execution/index.html#fundamental-types
Simon Atanasyan [Thu, 20 Jul 2017 20:34:18 +0000 (20:34 +0000)]
[CodeGen][mips] Support `long_call/far/near` attributes
This patch adds support for the `long_call`, `far`, and `near` attributes
for MIPS targets. The `long_call` and `far` attributes are synonyms. All
these attributes override `-mlong-calls` / `-mno-long-calls` command
line options for particular function.
This diff addresses FIXMEs in lib/Analysis/ScanfFormatString.cpp
for the case of ssize_t format specifier and adds tests.
In particular, this change enables Clang to emit a warning
on incorrect using of "%zd"/"%zn".
Peter Szecsi [Wed, 19 Jul 2017 23:50:00 +0000 (23:50 +0000)]
This feature allows the analyzer to consider loops to completely unroll. New
requirements/rules (for unrolling) can be added easily via ASTMatchers.
The current implementation is hidden behind a flag.
Right now the blocks which belong to an unrolled loop are marked by the
LoopVisitor which adds them to the ProgramState. Then whenever we encounter a
CFGBlock in the processCFGBlockEntrance which is marked then we skip its
investigating. That means, it won't be considered to be visited more than the
maximal bound for visiting since it won't be checked.
Hans Wennborg [Wed, 19 Jul 2017 12:31:01 +0000 (12:31 +0000)]
Revert r308441 "Recommit r308327: Add a warning for missing '#pragma pack (pop)' and suspicious uses of '#pragma pack' in included files"
This seems to have broken the sanitizer-x86_64-linux buildbot. Reverting until
it's fixed, especially since this landed just before the 5.0 branch.
> This commit adds a new -Wpragma-pack warning. It warns in the following cases:
>
> - When a translation unit is missing terminating #pragma pack (pop) directives.
> - When entering an included file if the current alignment value as determined
> by '#pragma pack' directives is different from the default alignment value.
> - When leaving an included file that changed the state of the current alignment
> value.
>
> rdar://10184173
>
> Differential Revision: https://reviews.llvm.org/D35484
Alex Lorenz [Wed, 19 Jul 2017 11:30:41 +0000 (11:30 +0000)]
Recommit r308327: Add a warning for missing '#pragma pack (pop)'
and suspicious uses of '#pragma pack' in included files
This commit adds a new -Wpragma-pack warning. It warns in the following cases:
- When a translation unit is missing terminating #pragma pack (pop) directives.
- When entering an included file if the current alignment value as determined
by '#pragma pack' directives is different from the default alignment value.
- When leaving an included file that changed the state of the current alignment
value.
[analyzer] Add annotation attribute to trust retain count implementation
Add support to the retain-count checker for an annotation indicating that a
function's implementation should be trusted by the retain count checker.
Functions with these attributes will not be inlined and the arguments will
be treating as escaping.
Adding this annotation avoids spurious diagnostics when the implementation of
a reference counting operation is visible but the analyzer can't reason
precisely about the ref count.
Petr Hosek [Wed, 19 Jul 2017 00:29:41 +0000 (00:29 +0000)]
[scan-build-py] Patch to fix "-analyzer-config" option
I noticed that when I use "-analyze-config" option in scan-build-py, it
behaves differently from original perl based scan-build.
For example, command:
$ scan-build -analyzer-config ipa=basic-inlining make
Will work without any issues on perl version of scan-build. But on
scan-build-py it will throw an error message "error reading
'ipa=basic-inlining'".
After debugging, it turns out that the scan-build-py does not put
"-analyzer-config" flag in front of the analyzer config flags (in this
case is the "ipa=basic-inlining") in the final clang command line. This
patch fixes this issue.
Adrian Prantl [Tue, 18 Jul 2017 23:58:34 +0000 (23:58 +0000)]
Debug Info: Set the MainFileName when generating -gmodules debug info for PCM.
Previously it was uninitialized and thus always defaulted to "<stdin>".
This is mostly a cosmetic change that helps making the debug info more readable.
Don't set TUScope to null when generating a module in incremental processing mode.
Summary: When in incremental processing mode, we should never set `TUScope` to a nullptr otherwise any future lookups fail. We already have similar checks in the rest of the code, but we never hit this one because so far we didn't try to generate a module from the AST that Cling generates.
Alex Lorenz [Tue, 18 Jul 2017 17:23:51 +0000 (17:23 +0000)]
Add a warning for missing '#pragma pack (pop)' and suspicious uses
of '#pragma pack' in included files
This commit adds a new -Wpragma-pack warning. It warns in the following cases:
- When a translation unit is missing terminating #pragma pack (pop) directives.
- When entering an included file if the current alignment value as determined
by '#pragma pack' directives is different from the default alignment value.
- When leaving an included file that changed the state of the current alignment
value.
Yaxun Liu [Tue, 18 Jul 2017 14:46:03 +0000 (14:46 +0000)]
CodeGen: Insert addr space cast for automatic/temp var at right position
The uses of alloca may be in different blocks other than the block containing the alloca.
Therefore if the alloca addr space is non-zero and it needs to be casted to default
address space, the cast needs to be inserted in the same BB as the alloca insted of
the current builder insert point since the current insert point may be in a different BB.
Martin Probst [Tue, 18 Jul 2017 14:00:19 +0000 (14:00 +0000)]
clang-format: [JS] Correctly format JavaScript imports with long module paths
Currently the `UnwrappedLineParser` fails to correctly unwrap JavaScript
imports where the module path is not on the same line as the `from` keyword.
For example:
import {A} from
'some/path/longer/than/column/limit/module.js';```
This causes issues when in the middle a list of imports because the formatter
thinks it has reached the end of the imports, and therefore will not sort any
imports lower in the list.
The formatter will, however, split the `from` keyword and the module path if
the path exceeds the column limit, which triggers the issue the next time the
file is formatted.
[CMake] Move CLANG_ENABLE_(ARCMT|OBJC_REWRITER|STATIC_ANALYZER) into clang/Config/config.h.
LLVM_ENABLE_MODULES is sensitive of -D. Move them into config.h.
FIXME: It'd be better that they are #cmakedefine01 rather than #cmakedefine.
(#if FOO rather than #if defined(FOO))
Then we can find missing #include "clang/Config/config.h" in the future.
Summary:
This patch prevents getCanonicalDecl returning nullptr in case it finds
a canonical TemplateDeclaration with no attached TemplatedDecl.
Found by running the indexer over a version of the standard library deep inside
a template metaprogramming mess.
projects / clang / log