David Majnemer [Thu, 5 Mar 2015 00:46:22 +0000 (00:46 +0000)]
MS ABI: Implement support for throwing a C++ exception
Throwing a C++ exception, under the MS ABI, is implemented using three
components:
- ThrowInfo structure which contains information like CV qualifiers,
what destructor to call and a pointer to the CatchableTypeArray.
- In a significant departure from the Itanium ABI, copying by-value
occurs in the runtime and not at the catch site. This means we need
to enumerate all possible types that this exception could be caught as
and encode the necessary information to convert from the exception
object's type to the catch handler's type. This includes complicated
derived to base conversions and the execution of copy-constructors.
N.B. This implementation doesn't support the execution of a
copy-constructor from within the runtime for now. Adding support for
that functionality is quite difficult due to things like default
argument expressions which may evaluate arbitrary code hiding in the
copy-constructor's parameters.
Reid Kleckner [Wed, 4 Mar 2015 19:24:16 +0000 (19:24 +0000)]
Fix test/CodeGen/builtins.c for platforms that don't lower sjlj
Opt in Win64 to supporting sjlj lowering. We have the backend lowering,
so I think this was just an oversight because WinX86_64TargetCodeGenInfo
doesn't inherit from X86_64TargetCodeGenInfo.
Toma Tabacu [Wed, 4 Mar 2015 14:24:25 +0000 (14:24 +0000)]
[IAS] Teach -cc1as about the 'target-abi' option.
Summary:
When using the IAS from clang, the 'target-abi' option gets passed to cc1as, but cc1as doesn't know about it and gives an "unknown argument" error.
This is fixed by adding the 'CC1AsOption' flag to the 'target-abi' option in CC1Options.td.
Seth Cantrell [Wed, 4 Mar 2015 03:12:10 +0000 (03:12 +0000)]
Add a format warning for "%p" with non-void* args
GCC -pedantic produces a format warning when the "%p" specifier is used with
arguments that are not void*. It's useful for portability to be able to
catch such warnings with clang as well. The warning is off by default in
both gcc and with this patch. This patch enables it either when extensions
are disabled with -pedantic, or with the specific flag -Wformat-pedantic.
The C99 and C11 specs do appear to require arguments corresponding to 'p'
specifiers to be void*: "If any argument is not the correct type for the
corresponding conversion specification, the behavior is undefined."
[7.19.6.1 p9], and of the 'p' format specifier "The argument shall be a
pointer to void." [7.19.6.1 p8]
Both printf and scanf format checking are covered.
Jacques Pienaar [Tue, 3 Mar 2015 23:58:09 +0000 (23:58 +0000)]
TypePrinter print __restrict if not in C99 mode
restrict is a keyword in C99 but not in C++ while clang accepts __restrict for C++ code. Modify the TypePrinter to print __restrict when not processing C99 code.
Printing restrict in C++ was problematic as printing the argument of
int f(int * __restrict a) { ... }
resulted in
int *restrict a
which is incorrect.
Anton Yartsev [Tue, 3 Mar 2015 22:58:46 +0000 (22:58 +0000)]
[analyzer] unix.Malloc: preserve AllocaRegion bound to __builtin_alloca().
Binding __builtin_alloca() return value to the symbolic value kills previous binding to a AllocaRegion established by the core.BuiltinFunctions checker. Other checkers may rely upon this information. Rollback handling of __builtin_alloca() to the way prior to r229850.
This test doesn't provide any value (it just checks that the frontend
produces exactly one compile unit), and it certainly isn't doing what
the comment says. Noticed via IRC review of my update to it in r231083.
Alexey Samsonov [Tue, 3 Mar 2015 22:15:35 +0000 (22:15 +0000)]
[UBSan] Split -fsanitize=shift into -fsanitize=shift-base and -fsanitize=shift-exponent.
-fsanitize=shift is now a group that includes both these checks, so
exisiting users should not be affected.
This change introduces two new UBSan kinds that sanitize only left-hand
side and right-hand side of shift operation. In practice, invalid
exponent value (negative or too large) tends to cause more portability
problems, including inconsistencies between different compilers, crashes
and inadequeate results on non-x86 architectures etc. That is,
-fsanitize=shift-exponent failures should generally be addressed first.
As a bonus, this change simplifies CodeGen implementation for emitting left
shift (separate checks for base and exponent are now merged by the
existing generic logic in EmitCheck()), and LLVM IR for these checks
(the number of basic blocks is reduced).
Bill Seurer [Tue, 3 Mar 2015 20:08:43 +0000 (20:08 +0000)]
[PowerPC]Activate "vector bool long long" (and alternate spellings) as a valid type for Altivec support for Power.
There are two test case updates for very basic testing. While I was editing cxx-altivec.cpp I also updated it to better match some other changes in altivec.c.
Note: "vector bool long" was not also added because its use is considered deprecated.
Dan Albert [Tue, 3 Mar 2015 18:24:57 +0000 (18:24 +0000)]
Don't force -pie for Android.
Summary:
There is no -no-pie flag that can override this, so making it default
to being on for Android means it is no longer possible to create
non-PIE executables on Android. While current versions of Android
support (and the most recent requires) PIE, ICS and earlier versions
of Android cannot run PIE executables, so this needs to be optional.
Hans Wennborg [Tue, 3 Mar 2015 17:30:50 +0000 (17:30 +0000)]
Migrate clang-format-vs plugin project to VS 2013
The plugin still works fine in versions starting from 2010,
but this was needed to make the project _build_ in VS 2013, which
is the blessed version for building LLVM projects these days.
[SDK modernizer]. Patch fixes driver's lack of
recognition of mernizer's -objcmt-migrate-property-dot-syntax
option with a new test in test/Driver. rdar://19994452
Manuel Klimek [Tue, 3 Mar 2015 14:54:25 +0000 (14:54 +0000)]
Make -Wuninitialized warn on pointer-to-member and comma operators.
`isTrackedVar` has been updated to also track records.
`DeclRefExpr`s appearing on the left side of a comma operator are
ignored, while those appearing on the right side are classified as
`Use`.
Daniel Jasper [Tue, 3 Mar 2015 13:59:49 +0000 (13:59 +0000)]
clang-format: Fix access to uninitialized memory.
With incomplete code, we aren't guaranteed to generated changes for
every token. In that case, we need to assume that even the very first
change can continue a preprocessor directive and initialize values
accordingly.
Daniel Jasper [Tue, 3 Mar 2015 10:02:53 +0000 (10:02 +0000)]
Revert r231008 (and dependent r231019).
As Chandler responded on the initial commit, just directly setting the
triple through -Xclang option to the driver creates havoc on other
platforms. The driver test should specifically go into test/Driver and
test the cc1 commandline itself.
David Majnemer [Tue, 3 Mar 2015 04:38:34 +0000 (04:38 +0000)]
Sema: Caught exception objects should be unqualified
The exception object should be unqualified. Using a qualified exception
object results in the wrong copy constructor getting called when the
catch handler executes.
Alexey Samsonov [Tue, 3 Mar 2015 00:14:32 +0000 (00:14 +0000)]
[Sanitizers] Remove duplication in sanitizer group definition. NFC.
There is no need to list sanitizers in both "UndefinedTrap" and
"Undefined" groups - it turns out using one group in a defintion
of another group "just works".
Reid Kleckner [Mon, 2 Mar 2015 22:42:58 +0000 (22:42 +0000)]
Remove shell requirements from tests that use 'cd'
Modules and Tooling tests in particular tend to want to change the cwd,
so we were missing test coverage in this area on Windows. It should now
be easier to write such portable tests.
Sanjoy Das [Mon, 2 Mar 2015 21:47:47 +0000 (21:47 +0000)]
Fix test case.
r230921 broke backend-optimization-failure.cpp: after
r230921, LLVM no longer emits an expression to compute 'Length - 1'
and this perturbs LoopSimplify enough to emit the warning on line 10
instead of line 9. This is a review request to fix the test case once
I re-land r230921.
Matthias Braun [Mon, 2 Mar 2015 20:44:09 +0000 (20:44 +0000)]
Improve robustness of dependency-generation-crash.c test.
The test wants to provoke a failure when opening the output file.
Using chmod 0 on the output file does not work reliably on all
filesystems or when running the test as root.
Change the test to use a nonexistant directory instead.
Add -frtti and -fexceptions to tests that assume these are on.
Summary:
We now have targets that don't enable rtti/exceptions by default, and the
ASTMatchers tests are assuming that these features are on (e.g: They use
dynamic_cast or try).
Benjamin Kramer [Mon, 2 Mar 2015 16:09:24 +0000 (16:09 +0000)]
CodeGen: Fix passing of classes with only one AVX vector member in AVX registers
isSingleElementStruct was a bit too tight in its definition of struct
so we got a mismatch between classify() and the actual code generation.
To make matters worse the code in GetByteVectorType still defaulted to
<2 x double> if it encountered a type it didn't know, making this a
silent miscompilation (PR22753).
Completely remove the "preferred type" stuff from GetByteVectorType and
make it fail an assertion if someone tries to use it with a type not
suitable for a vector register.
Richard Smith [Sat, 28 Feb 2015 03:09:52 +0000 (03:09 +0000)]
Give better diagnostics when -fmodule-file= finds a bad file: if the file is
found indirectly, explain how we got there, and distinguish between 'file not
found' and 'file found but invalid'.
Richard Smith [Sat, 28 Feb 2015 01:01:56 +0000 (01:01 +0000)]
Rework our handling of key functions. We used to track a complete list of all
dynamic classes in the translation unit and check whether each one's key
function is defined when we got to the end of the TU (and when we got to the
end of each module). This is really terrible for modules performance, since it
causes unnecessary deserialization of every dynamic class in every compilation.
We now use a much simpler (and, in a modules build, vastly more efficient)
system: when we see an out-of-line definition of a virtual function, we check
whether that function was in fact its class's key function. (If so, we need to
emit the vtable.)
DebugInfo: hoist definition into global context when needed
When generating debug info for a static inline member which is initialized for
the DLLExport storage class, hoist the definition into a non-composite type
context. Otherwise, we would trigger an assertion when generating the DIE for
the associated global value as the debug context has a type association. This
addresses PR22669.
Thanks to David Blakie for help in coming up with a solution to this!
Richard Smith [Fri, 27 Feb 2015 23:05:10 +0000 (23:05 +0000)]
[modules] When writing out a list of specializations for a template, if we have
undeserialized specializations (because we merged an imported declaration of
the same template since we last added one), don't bother reading in the
specializations themselves just so we can write out their IDs again.
Daniel Jasper [Fri, 27 Feb 2015 08:16:32 +0000 (08:16 +0000)]
Revert r230717 (and subsequent r230720).
The tests keeps failing on build bots:
http://lab.llvm.org:8080/green/job/clang-stage2-configure-Rlto_check/2355/testReport/junit/Clang/CodeGen/exceptions_seh_leave_c/
Alexey Bataev [Fri, 27 Feb 2015 06:33:30 +0000 (06:33 +0000)]
[OPENMP] Codegen for "#pragma omp atomic write"
For global reg lvalue - use regular store through global register.
For simple lvalue - use simple atomic store.
For bitfields, vector element, extended vector elements - the original value of the whole storage (for vector elements) or of some aligned value (for bitfields) is atomically read, the part of this value for the given lvalue is modified and then use atomic compare-and-exchange operation to try to atomically write modified value (if it was not modified).
Also, changes in this patch fix the bug for '#pragma omp atomic read' applied to extended vector elements.
Differential Revision: http://reviews.llvm.org/D7369
Richard Smith [Fri, 27 Feb 2015 01:57:00 +0000 (01:57 +0000)]
[modules] For an inheriting constructor, the inherited constructor is stored in
a map keyed off the canonical declaration. Don't try to set it if we're loading
some non-canonical merged declaration.
Richard Smith [Fri, 27 Feb 2015 00:25:58 +0000 (00:25 +0000)]
[modules] When loading in multiple canonical definitions of a template,
accumulate the set of specializations rather than overwriting one list
with another.
Nico Weber [Thu, 26 Feb 2015 22:34:33 +0000 (22:34 +0000)]
Don't crash on leaving nested __finally blocks through an EH edge.
The __finally emission block tries to be clever by removing unused continuation
edges if there's an unconditional jump out of the __finally block. With
exception edges, the EH continuation edge isn't always unused though and we'd
crash in a few places.
Just don't be clever. That makes the IR for __finally blocks a bit longer in
some cases (hence small and behavior-preserving changes to existing tests), but
it makes no difference in general and it fixes the last crash from PR22553.