Sam McCall [Tue, 18 Sep 2018 08:40:41 +0000 (08:40 +0000)]
[CodeComplete] Add completions for filenames in #include directives.
Summary:
The dir component ("somedir" in #include <somedir/fo...>) is considered fixed.
We append "foo" to each directory on the include path, and then list its files.
Completions are of the forms:
#include <somedir/fo^
foo.h>
fox/
The filter is set to the filename part ("fo"), so fuzzy matching can be
applied to the filename only.
No fancy scoring/priorities are set, and no information is added to
CodeCompleteResult to make smart scoring possible. Could be in future.
Thread safety analysis: Run more tests with capability attributes [NFC]
Summary:
We run the tests for -Wthread-safety-{negative,verbose} with the new
attributes as well as the old ones. Also put the macros in a header so
that we don't have to copy them all around.
The warn-thread-safety-parsing.cpp test checks for warnings depending on
the actual attribute name, so it can't undergo the same treatment.
George Karpenkov [Mon, 17 Sep 2018 20:46:53 +0000 (20:46 +0000)]
[analyzer] ExplodedGraph printing fixes
Fixes a number of issues:
- Global variables are not used for communication
- Trait should be defined on a graph, not on a node
- Defining the trait on a graph allows us to use a correct allocator,
no longer crashing while printing trimmed graphs
Shuai Wang [Mon, 17 Sep 2018 20:10:56 +0000 (20:10 +0000)]
[analyzer] Treat std::{move,forward} as casts in ExprMutationAnalyzer.
Summary:
This is a follow up of D52008 and should make the analyzer being able to handle perfect forwardings in real world cases where forwardings are done through multiple layers of function calls with `std::forward`.
Summary:
The init expression of a VarDecl is overwritten in the "To" context if we
import a VarDecl without an init expression (and with a definition). Please
refer to the added tests, especially InitAndDefinitionAreInDifferentTUs. This
patch fixes the malfunction by importing the whole Decl chain similarly as we
did that in case of FunctionDecls. We handle the init expression similarly to
a definition, alas only one init expression will be in the merged ast.
Summary:
Merged the recently added `err_attribute_argument_negative` diagnostic
with existing `err_attribute_requires_positive_integer` diagnostic:
the former allows only strictly positive integer, while the latter
also allows zero.
scan-build: Using '/usr/lib/llvm-7/bin/clang' for static analysis
scan-build: Emitting reports for this run to '/tmp/scan-build-2018-09-16-214531-8410-1'.
foo.c:3:3: warning: Value stored to 'x' is never read
x = 1; // warn
^ ~
1 warning generated.
scan-build: File '/tmp/foo.c' deleted: part of an ignored directory.
scan-build: 0 bugs found.
```
Kelvin Li [Sat, 15 Sep 2018 13:54:15 +0000 (13:54 +0000)]
[OPENMP] Move OMPClauseReader/Writer classes to ASTReader/Writer (NFC)
Move declarations for OMPClauseReader, OMPClauseWriter to ASTReader.h
and ASTWriter.h and move implementation to ASTReader.cpp and
ASTWriter.cpp. This change helps generalize the serialization of
OpenMP clauses and will be used in the future implementation of new
OpenMP directives (e.g. requires).
Shuai Wang [Fri, 14 Sep 2018 20:07:18 +0000 (20:07 +0000)]
[analyzer] Handle forwarding reference better in ExprMutationAnalyzer.
Summary:
We used to treat an `Expr` mutated whenever it's passed as non-const
reference argument to a function. This results in false positives in
cases like this:
```
int x;
std::vector<int> v;
v.emplace_back(x); // `x` is passed as non-const reference to `emplace_back`
```
In theory the false positives can be suppressed with
`v.emplace_back(std::as_const(x))` but that's considered overly verbose,
inconsistent with existing code and spammy as diags.
This diff handles such cases by following into the function definition
and see whether the argument is mutated inside.
Hans Wennborg [Fri, 14 Sep 2018 15:18:30 +0000 (15:18 +0000)]
[clang-cl] Fix PR38934: failing to dllexport class template member w/ explicit instantiation and PCH
The code in ASTContext::DeclMustBeEmitted was supposed to handle this,
but didn't take into account that synthesized members such as operator=
might not get marked as template specializations, because they're
synthesized on the instantiation directly when handling the class-level
dllexport attribute.
Louis Dionne [Fri, 14 Sep 2018 14:07:16 +0000 (14:07 +0000)]
[clang] Make sure attributes on member classes are applied properly
Summary:
Attributes on member classes of class templates and member class templates
of class templates are not currently instantiated. This was discovered by
Richard Smith here:
This commit makes sure that attributes are instantiated properly. This
commit does not fix the broken behavior for member partial and explicit
specializations of class templates.
Sam McCall [Fri, 14 Sep 2018 12:47:38 +0000 (12:47 +0000)]
[VFS] vfs::directory_iterator yields path and file type instead of full Status
Summary:
Most callers I can find are using only `getName()`. Type is used by the
recursive iterator.
Now we don't have to call stat() on every listed file (on most platforms).
Exceptions are e.g. Solaris where readdir() doesn't include type information.
On those platforms we'll still stat() - see D51918.
The result is significantly faster (stat() can be slow).
My motivation: this may allow us to improve clang IO on large TUs with long
include search paths. Caching readdir() results may allow us to skip many stat()
and open() operations on nonexistent files.
[Driver] Fix missing MultiArch include dir on powerpcspe
On powerpc-linux-gnuspe, the header files are located in their
own include directory named /usr/lib/powerpc-linux-gnuspe,
so add this directory to PPCMultiarchIncludeDirs.
Sam McCall [Fri, 14 Sep 2018 12:24:09 +0000 (12:24 +0000)]
[Tooling] JSONCompilationDatabasePlugin infers compile commands for missing files
Summary:
See the existing InterpolatingCompilationDatabase for details on how this works.
We've been using this in clangd for a while, the heuristics seem to work well.
[analyzer][UninitializedObjectChecker] New flag to ignore records based on it's fields
Based on a suggestion from @george.karpenkov.
In some cases, structs are used as unions with a help of a tag/kind field.
This patch adds a new string flag (a pattern), that is matched against the
fields of a record, and should a match be found, the entire record is ignored.
For more info refer to http://lists.llvm.org/pipermail/cfe-dev/2018-August/058906.html
and to the responses to that, especially http://lists.llvm.org/pipermail/cfe-dev/2018-August/059215.html.
Since I plan to add a number of new flags, it made sense to encapsulate
them in a new struct, in order not to pollute FindUninitializedFields's
constructor with new boolean options with super long names.
This revision practically reverts D50508, since FindUninitializedFields
now accesses the pedantic flag anyways.
Some of the comments are incorrect, imprecise, or simply nonexistent.
Since I have a better grasp on how the analyzer works, it makes sense
to update most of them in a single swoop.
I tried not to flood the code with comments too much, this amount
feels just right to me.
Summary:
Before this change, we only emit the XRay attributes in LLVM IR when the
-fxray-instrument flag is provided. This may cause issues with thinlto
when the final binary is being built/linked with -fxray-instrument, and
the constitutent LLVM IR gets re-lowered with xray instrumentation.
With this change, we can honour the "never-instrument "attributes
provided in the source code and preserve those in the IR. This way, even
in thinlto builds, we retain the attributes which say whether functions
should never be XRay instrumented.
Richard Trieu [Fri, 14 Sep 2018 01:15:28 +0000 (01:15 +0000)]
[ODRHash] Fix early exit that skipped code.
There is a bit of code at the end of AddDeclaration that should be run on
every exit of the function. However, there was an early exit beforehand
that could be triggered, which causes a small amount of data to skip the
hashing, leading to false positive mismatch. Use a separate function so
that this code is always run.
[Sema] Remove location from implicit capture init expr
A lambda's closure is initialized when the lambda is declared. For
implicit captures, the initialization code emitted from EmitLambdaExpr
references source locations *within the lambda body* in the function
containing the lambda. This results in a poor debugging experience: we
step to the line containing the lambda, then into lambda, out again,
over and over, until every capture's field is initialized.
To improve stepping behavior, assign the starting location of the lambda
to expressions which initialize an implicit capture within it.
Richard Smith [Thu, 13 Sep 2018 21:10:08 +0000 (21:10 +0000)]
Diagnose likely typos in #include directives.
Summary:
When someone writes
#include "<some_file>"
or
#include " some_file "
the compiler returns "file not fuond..." with fonts and quotes that may
make it hard to see there are excess quotes or surprising bytes in the
filename. Assuming that files are usually logically named and start and
end with an alphanumeric character, we can check for the file's
existence by stripping the non-alphanumeric leading or trailing
characters. If the file is found, emit a non-fatal error with a
FixItHint.
Stephen Hines [Thu, 13 Sep 2018 19:50:02 +0000 (19:50 +0000)]
Support -fno-omit-frame-pointer with -pg.
Summary:
Previously, any instance of -fomit-frame-pointer would make it such that
-pg was an invalid flag combination. If -fno-omit-frame-pointer is
passed later on the command line (such that it actually takes effect),
-pg should be allowed.
Erich Keane [Thu, 13 Sep 2018 16:58:24 +0000 (16:58 +0000)]
[NFC]Refactor MultiVersion Resolver Emission to combine types
Previously, both types (plus the future target-clones) of
multiversioning had a separate ResolverOption structure and emission
function. This patch combines the two, at the expense of a slightly
more expensive sorting function.
[OPENMP] Fix PR38903: Crash on instantiation of the non-dependent
declare reduction.
If the declare reduction construct with the non-dependent type is
defined in the template construct, the compiler might crash on the
template instantition. Reworked the whole instantiation scheme for the
declare reduction constructs to fix this problem correctly.
Oliver Stannard [Thu, 13 Sep 2018 10:25:36 +0000 (10:25 +0000)]
[AArch64] Enable return address signing for static ctors
Functions generated by clang and included in the .init_array section (such as
static constructors) do not follow the usual code path for adding
target-specific function attributes, so we have to add the return address
signing attribute here too, as is currently done for the sanitisers.
Tri Vo [Wed, 12 Sep 2018 23:45:04 +0000 (23:45 +0000)]
[AArch64] Support reserving x1-7 registers.
Summary: Reserving registers x1-7 is used to support CONFIG_ARM64_LSE_ATOMICS in Linux kernel. This change adds support for reserving registers x1 through x7.
Richard Smith [Wed, 12 Sep 2018 23:37:00 +0000 (23:37 +0000)]
Track definition merging on the canonical declaration even when local
submodule visibility is disabled.
Attempting to pick a specific declaration to make visible when the
module containing the merged declaration becomes visible is error-prone,
as we don't yet know which declaration we'll choose to be the definition
when we are informed of the merging.
This reinstates r342019, reverted in r342020. The regression previously
observed after this commit was fixed in r342096.
[RISCV] Explicitly set an empty --sysroot in the test
In rL341655 we added additional behaviour to the Driver for riscv32-unknown-elf
when the sysroot is empty.
The new tests that check the new behaviour expect that the absence of --sysroot
in the command-line implies that the sysroot empty. This doesn't hold if clang
is built with a non-empty DEFAULT_SYSROOT in cmake. When this is the case, this
test fails.
Since the new behaviour is triggered when the sysroot is empty, pass an empty
--sysroot to avoid using the default (if any).
David Green [Wed, 12 Sep 2018 14:09:06 +0000 (14:09 +0000)]
[CodeGen] Align rtti and vtable data
Previously the alignment on the newly created rtti/typeinfo data was largely
not set, meaning that DataLayout::getPreferredAlignment was free to overalign
it to 16 bytes. This causes unnecessary code bloat.
[CodeGen][ARM] Coerce FP16 vectors to integer vectors when needed
Summary:
On targets that do not support FP16 natively LLVM currently legalizes
vectors of FP16 values by scalarizing them and promoting to FP32. This
causes problems for the following code:
According to the AAPCS (appendix A.2) float16x4_t is a containerized
vector fundamental type, so 'foo' expects that the 4 16-bit FP values
are packed into 2 32-bit registers, but instead bar promotes them to
4 single precision values.
Since we already handle scalar FP16 values in the frontend by
bitcasting them to/from integers, this patch adds similar handling for
vector types and homogeneous FP16 vector aggregates.
One existing test required some adjustments because we now generate
more bitcasts (so the patch changes the test to target a machine with
native FP16 support).
Richard Smith [Wed, 12 Sep 2018 02:13:48 +0000 (02:13 +0000)]
Track definition merging on the canonical declaration even when local
submodule visibility is disabled.
Attempting to pick a specific declaration to make visible when the
module containing the merged declaration becomes visible is error-prone,
as we don't yet know which declaration we'll choose to be the definition
when we are informed of the merging.
Richard Smith [Wed, 12 Sep 2018 02:13:47 +0000 (02:13 +0000)]
Consistently create a new declaration when merging a pre-existing but
hidden definition with a would-be-parsed redefinition.
This permits a bunch of cleanups. In particular, we no longer need to
take merged definitions into account when checking declaration
visibility, only when checking definition visibility, which makes
certain visibility checks take linear instead of quadratic time.
We could also now remove the UPD_DECL_EXPORTED update record and track
on each declaration whether it was demoted from a definition (as we
already do for variables), but I'm not doing that in this patch to keep
the changes here simpler.
Shuai Wang [Tue, 11 Sep 2018 21:13:20 +0000 (21:13 +0000)]
[analyzer] Add ExprMutationAnalyzer
Summary:
This is 1/2 of moving ExprMutationAnalyzer from clangtidy to
clang/Analysis.
This diff along simply copies the ExprMutationAnalyzer over with trivial
modifications (e.g. include path, namespace)
2/2 will migrate existing usage of ExprMutationAnalyzer and remove the
original copy inside clangtidy.
Mike Rice [Tue, 11 Sep 2018 17:10:44 +0000 (17:10 +0000)]
[clang-cl, PCH] Support for /Yc and /Yu without filename and #pragma hdrstop
With clang-cl, when the user specifies /Yc or /Yu without a filename
the compiler uses a #pragma hdrstop in the main source file to
determine the end of the PCH. If a header is specified with /Yc or
/Yu #pragma hdrstop has no effect.
The optional #pragma hdrstop filename argument is not yet supported.
Jonas Toth [Tue, 11 Sep 2018 16:09:19 +0000 (16:09 +0000)]
[ASTMatchers] add three matchers for dependent expressions
Summary:
The new matchers can be used to check if an expression is type-, value- or instantiation-dependent
in a templated context.
These matchers are used in a clang-tidy check and generally useful as the
problem of unresolved templates occurs more often in clang-tidy and they
provide an easy way to check for this issue.
[CodeCompletion] Enable signature help when initializing class/struct/union members.
Summary:
Factors out member decleration gathering and uses it in parsing to call signature
help. Doesn't support signature help for base class constructors, the code was too
coupled with diagnostic handling, but still can be factored out but just needs
more afford.
[OPENMP] Simplified checks for declarations in declare target regions.
Sema analysis should not mark functions as an implicit declare target,
it may break codegen. Simplified semantic analysis and removed extra
code for implicit declare target functions.
Summary:
And add an option to disable this behavior. The option is only used in
AllTUsExecutor to avoid races when running concurrently on multiple
threads.
projects / clang / log