Justin Lebar [Wed, 24 Feb 2016 21:49:31 +0000 (21:49 +0000)]
[CUDA] Add hack so code which includes "curand.h" doesn't break.
Summary:
curand.h includes curand_mtgp32_kernel.h. In host mode, this header
redefines threadIdx and blockDim, giving them their "proper" types of
uint3 and dim3, respectively.
clang has its own plan for these variables -- their types are magic
builtin classes. So these redefinitions are incompatible.
As a hack, we force-include the offending CUDA header and use #defines
to get the right types for threadIdx and blockDim.
Justin Lebar [Wed, 24 Feb 2016 21:49:28 +0000 (21:49 +0000)]
Bail on compilation as soon as a job fails.
Summary:
(Re-land of r260448, which was reverted in r260522 due to a test failure
in Driver/output-file-cleanup.c that only showed up in fresh builds.)
Previously we attempted to be smart; if one job failed, we'd run all
jobs that didn't depend on the failing job.
Problem is, this doesn't work well for e.g. CUDA compilation without
-save-temps. In this case, the device-side and host-side Assemble
actions (which actually are responsible for preprocess, compile,
backend, and assemble, since we're not saving temps) are necessarily
distinct. So our clever heuristic doesn't help us, and we repeat every
error message once for host and once for each device arch.
The main effect of this change, other than fixing CUDA, is that if you
pass multiple cc files to one instance of clang and you get a compile
error, we'll stop when the first cc1 job fails.
Reviewers: echristo
Subscribers: cfe-commits, jhen, echristo, tra, rafael
Nico Weber [Wed, 24 Feb 2016 20:58:14 +0000 (20:58 +0000)]
Fix rejects-valid caused by r261297.
r261297 called hasUserProvidedDefaultConstructor() to check if defining a
const object is ok. This is incorrect for this example:
struct X { template<typename ...T> X(T...); int n; };
const X x; // formerly OK, now bogus error
Instead, track if a class has a defaulted default constructor, and disallow
a const object for classes that either have defaulted default constructors or
if they need an implicit constructor.
Bug report and fix approach by Richard Smith, thanks!
James Y Knight [Wed, 24 Feb 2016 02:59:33 +0000 (02:59 +0000)]
Default vaarg lowering should support indirect struct types.
Fixes PR11517 for SPARC.
On most targets, clang lowers va_arg itself, eschewing the use of the
llvm vaarg instruction. This is necessary (at least for now) as the type
argument to the vaarg instruction cannot represent all the ABI
information that is needed to support complex calling conventions.
However, on targets with a simpler varrags ABIs, the LLVM instruction
can work just fine, and clang can simply lower to it. Unfortunately,
even on such targets, vaarg with a struct argument would fail, because
the default lowering to vaarg was naive: it didn't take into account the
ABI attribute computed by classifyArgumentType. In particular, for the
DefaultABIInfo, structs are supposed to be passed indirectly and so
llvm's vaarg instruction should be emitted with a pointer argument.
Now, vaarg instruction emission is able to use computed ABIArgInfo for
the provided argument type, which allows the default ABI support to work
for structs too.
I haven't touched the EmitVAArg implementation for PPC32_SVR4 or XCore,
although I believe both are now redundant, and could be switched over to
use the default implementation as well.
Devin Coughlin [Tue, 23 Feb 2016 22:26:04 +0000 (22:26 +0000)]
[analyzer] Find ObjC 'self' decl even when block captures local named 'self'.
When looking up the 'self' decl in block captures, make sure to find the actual
self declaration even when the block captures a local variable named 'self'.
Chris Bieneman [Tue, 23 Feb 2016 20:33:15 +0000 (20:33 +0000)]
[CMake] Add install-clang-format target by migrating to add_clang_tool
This change migrates clang-format to add_clang_tool which makes a component-based install target. To support component-based installation the extra installed scripts all need to have the "clang-format" component too.
Nico Weber [Tue, 23 Feb 2016 19:30:43 +0000 (19:30 +0000)]
Rename Action::begin() to Action::input_begin().
Also introduce inputs() that reutnrs an llvm::iterator_range.
Iterating over A->inputs() is much less mysterious than
iterating over *A. No intended behavior change.
Adrian Prantl [Tue, 23 Feb 2016 19:30:08 +0000 (19:30 +0000)]
Reapply r261657.
Remove an unnecessary workaround introduced in r259975. (NFC)
Now that LLVM r259973 allows replacing a temporary type with another
temporary we can rely on the original implementation.
It is possible for enums to be created as part of
their own declcontext. In this case a FwdDecl will be created
twice. This doesn't cause a problem because both FwdDecls are
entered into the ReplaceMap: finalize() will replace the first
FwdDecl with the second and then replace the second with
complete type.
Aaron Ballman [Tue, 23 Feb 2016 18:55:15 +0000 (18:55 +0000)]
Amends r252104 to evaluate the controlling expression in an unevaluated context. This eliminates false-positive diagnostics about null pointer dereferences (etc) in the controlling expression.
Adrian Prantl [Tue, 23 Feb 2016 17:13:47 +0000 (17:13 +0000)]
Remove an unnecessary workaround introduced in r259975. (NFC)
Now that LLVM r259973 allows replacing a temporary type with another
temporary we can rely on the original implementation.
It is possible for enums to be created as part of
their own declcontext. In this case a FwdDecl will be created
twice. This doesn't cause a problem because both FwdDecls are
entered into the ReplaceMap: finalize() will replace the first
FwdDecl with the second and then replace the second with
complete type.
Gabor Horvath [Tue, 23 Feb 2016 12:34:39 +0000 (12:34 +0000)]
[analyzer] Improve pointer arithmetic checker.
This patch is intended to improve pointer arithmetic checker.
From now on it only warns when the pointer arithmetic is likely to cause an
error. For example when the pointer points to a single object, or an array of
derived types.
Lex: Return "" when HeaderMap::lookupFilename fails
Change getString() to return Optional<StringRef>, and change
lookupFilename() to return an empty string if either one of the prefix
and suffix can't be found.
This is a more robust follow-up to r261461, but it's still not entirely
satisfactory. Ideally we'd report that the header map is corrupt;
perhaps something for a follow-up.
Lex: Check for 0 buckets on header map construction
Switch to using `isPowerOf2_32()` to check whether the buckets are a
power of two, and as a side benefit reject loading a header map with no
buckets. This is a follow-up to r261448.
The VFS overlay mapping between virtual paths and real paths is done through
the 'external-contents' entries in YAML files, which contains hardcoded paths
to the real files.
When a module compilation crashes, headers are dumped into <name>.cache/vfs
directory and are mapped via the <name>.cache/vfs/vfs.yaml. The script
generated for reproduction uses -ivfsoverlay pointing to file to gather the
mapping between virtual paths and files inside <name>.cache/vfs. Currently, we
are only capable of reproducing such crashes in the same machine as they
happen, because of the hardcoded paths in 'external-contents'.
To be able to reproduce a crash in another machine, this patch introduces a new
option in the VFS yaml file called 'overlay-relative'. When it's equal to
'true' it means that the provided path to the YAML file through the
-ivfsoverlay option should also be used to prefix the final path for every
'external-contents'.
Example, given the invocation snippet "... -ivfsoverlay
<name>.cache/vfs/vfs.yaml" and the following entry in the yaml file:
Here, a file manager request for virtual "/usr/include/stdio.h", that will map
into real path "/<absolute_path_to>/<name>.cache/vfs/usr/include/stdio.h.
This is a useful feature for debugging module crashes in machines other than
the one where the error happened.
Handle ".", ".." and "./" with trailing slashes while collecting files
to be dumped into the vfs overlay directory.
Include the support for symlinks into components. Given the path:
/install-dir/bin/../lib/clang/3.8.0/include/altivec.h, if "bin"
component is a symlink, it's not safe to use `path::remove_dots` here,
and `realpath` is used to get the right answer. Since `realpath`
is expensive, we only do it at collecting time (which only happens
during the crash reproducer) and cache the base directory for fast lookups.
Overall, this makes the input to the VFS YAML file to be canonicalized
to never contain traversal components.
Devin Coughlin [Mon, 22 Feb 2016 17:56:24 +0000 (17:56 +0000)]
[analyzer] Detect duplicate [super dealloc] calls
Add an alpha path checker that warns about duplicate calls to [super dealloc].
This will form the foundation of a checker that will detect uses of
'self' after calling [super dealloc].
We gave a VBTable dllimport storage class and external linkage while
also providing an initializer. An initializer is only valid if the
VBTable has available_externally linkage. Fix this by setting the
linkage to available_externally in situ while generating the
initializer.
Anton Yartsev [Sun, 21 Feb 2016 17:04:26 +0000 (17:04 +0000)]
[analyzer][scan-build] Non-existing directory for scan-build output.
Makes scan-build successfully accept non-existing output directories provided via "-o" option. The directory is created in this case. This behavior is conforming to the old perl scan-build implementation.
(http://reviews.llvm.org/D17091)
Lex: Never overflow the file in HeaderMap::lookupFilename()
If a header map file is corrupt, the strings in the string table may not
be null-terminated. The logic here previously relied on `MemoryBuffer`
always being null-terminated, but this isn't actually guaranteed by the
class AFAICT. Moreover, we're seeing a lot of crash traces at calls to
`strlen()` inside of `lookupFilename()`, so something is going wrong
there.
Instead, use `strnlen()` to get the length, and check for corruption.
Also remove code paths that could call `StringRef(nullptr)`. r261459
made these rather obvious (although they'd been there all along).
Lex: Check whether the header map buffer has space for the buckets
Check up front whether the header map buffer has space for all of its
declared buckets.
There was already a check in `getBucket()`, but it had UB (comparing
pointers that were outside of objects in the error path) and was
insufficient (only checking for a single byte of the relevant bucket).
I fixed the check, moved it to `checkHeader()`, and left a fixed version
behind as an assertion.
If the number of buckets is not a power of two, immediately recognize
the header map as corrupt, rather than waiting for the first lookup. I
converted the later check to an assert.
Split the implementation of `HeaderMap` into `HeaderMapImpl` so that we
can write unit tests that don't depend on the `FileManager`, and then
write a few tests that cover the types of corrupt header maps already
detected.
This also moves type and constant definitions from HeaderMap.cpp to
HeaderMapTypes.h so that the test can access them.
Roman Divacky [Sat, 20 Feb 2016 08:31:24 +0000 (08:31 +0000)]
Fix handling of vaargs on PPC32 when going from regsave to overflow.
It can happen that when we only have 1 more register left in the regsave
area we need to store a value bigger than 1 register and therefore we
go to the overflow area. In this case we have to leave the last slot
in the regsave area unused and keep using overflow area. Do this
by storing a limit value to the used register counter in the overflow block.
Issue diagnosed by and solution tested by Mark Millard!
Richard Smith [Fri, 19 Feb 2016 22:43:58 +0000 (22:43 +0000)]
[modules] Do less scanning of macro definition chains when computing the set of
exported module macros outside local submodule visibility mode. Related to
PR24667.
Richard Smith [Fri, 19 Feb 2016 22:25:36 +0000 (22:25 +0000)]
[modules] Flatten -fmodule-name= and -fmodule-implementation-of= into a single
option. Previously these options could both be used to specify that you were
compiling the implementation file of a module, with a different set of minor
bugs in each case.
This change removes -fmodule-implementation-of, and instead tracks a flag to
determine whether we're currently building a module. -fmodule-name now behaves
the same way that -fmodule-implementation-of previously did.
Alexey Bataev [Fri, 19 Feb 2016 11:23:28 +0000 (11:23 +0000)]
pr26544: Bitfield layout with pragma pack and attributes "packed" and
"aligned", by Vladimir Yakovlev
Fix clang/gcc incompatibility of bitfields layout in the presence of
pragma packed and attributes aligned and packed.
Differential Revision: http://reviews.llvm.org/D17023
JF Bastien [Fri, 19 Feb 2016 06:54:45 +0000 (06:54 +0000)]
ARM: fix VFP asm constraints
Summary:
Rich Felker was sad that clang used 'w' and 'P' for VFP constraints when GCC documents them as 't' and 'w':
https://gcc.gnu.org/onlinedocs/gcc/Machine-Constraints.html
This was added way back in 2008:
http://lists.llvm.org/pipermail/cfe-commits/Week-of-Mon-20080421/005393.html
Nico Weber [Fri, 19 Feb 2016 01:52:46 +0000 (01:52 +0000)]
Implement the likely resolution of core issue 253.
C++11 requires const objects to have a user-provided constructor, even for
classes without any fields. DR 253 relaxes this to say "If the implicit default
constructor initializes all subobjects, no initializer should be required."
clang is currently the only compiler that implements this C++11 rule, and e.g.
libstdc++ relies on something like DR 253 to compile in newer versions. This
change makes it possible to build code that says `const vector<int> v;' again
when using libstdc++5.2 and _GLIBCXX_DEBUG
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60284).
Devin Coughlin [Fri, 19 Feb 2016 01:35:10 +0000 (01:35 +0000)]
[analyzer] Add checker callback for beginning of function.
Add a checker callback that is called when the analyzer starts analyzing a
function either at the top level or when inlined. This will be used by a
follow-on patch making the DeallocChecker path sensitive.
Reid Kleckner [Fri, 19 Feb 2016 00:56:56 +0000 (00:56 +0000)]
Add call to find_package to load LLVM dependencies
ClangConfig requires LLVMConfig, so add find_package call in
ClangConfig so find_package(clang REQUIRED CONFIG) will just work. This
makes it easier for cmake based projects to use clang, e.g., tools using
ClangTooling.
Richard Trieu [Thu, 18 Feb 2016 23:58:40 +0000 (23:58 +0000)]
Add -Wcomma warning to Clang.
-Wcomma will detect and warn on most uses of the builtin comma operator. It
currently whitelists the first and third statements of the for-loop. For other
cases, the warning can be silenced by casting the first operand of the comma
operator to void.
projects / clang / log