This changes the CompilerInstance::createOutputFile function to return
a std::unique_ptr<llvm::raw_ostream>, rather than an llvm::raw_ostream
implicitly owned by the CompilerInstance. This in most cases required that
I move ownership of the output stream to the relevant ASTConsumer.
The motivation for this change is to allow BackendConsumer to be a client
of interfaces such as D20268 which take ownership of the output stream.
Differential Revision: http://reviews.llvm.org/D21537
llvm-svn: 275507
Add an option to opt into compiler-rt instead of libgcc. This option defaults
to OFF to avoid a behaviour change. It is not possible to mix and match
different runtime libraries. Disabling this requires that libc++ is built
accordingly. This knob is particularly useful for targets that are GCC by
default (i.e. Linux).
llvm-svn: 275505
If a subtarget has both ZCZeroing and CustomCheapAsMoveHandling features (now
only Kryo has both), set COPY (W|X)ZR isAsCheapAsAMove.
Differential Revision: http://reviews.llvm.org/D22360
llvm-svn: 275503
For a fully inlined call chain like a -> b -> c -> d, we were emitting
line info for 'd' 3 separate times: once for d's actual InlineSite line
table, and twice for 'b' and 'c'. This is particularly inefficient when
all these functions are in different headers, because now we need to
encode the file change. Windbg was coping with our suboptimal output, so
this should not be noticeable from the debugger.
llvm-svn: 275502
We don't need to print any of the special __mh_*_header symbols when
disassembling. Since they point at the beginning of the segment (not where the
actual code is) they're pretty misleading.
Should also fix lld bots.
llvm-svn: 275498
This improves the situation discussed in D19228 where we were forcing VPERMPD/VPERMQ where VPERM2F128/VPERM2I128 would have been better.
This was incorrectly reverted in rL275421 during triage of PR28552.
llvm-svn: 275497
There is not enough padding in front of memchr(), and, the first 6 bytes
contains a branch instruction. Basically the current interception will
not work on memchr().
It was disabled before, but was missing the part to disable it for
INTERCEPT_LIBRARY_FUNCTION.
Patch by Wei Wang
Differential Revision: https://reviews.llvm.org/D22371
llvm-svn: 275494
Summary:
Both test have the same command-line.
The second test is missing the /GS-.
Keep in mind that /GS is on by default.
Reviewers: rnk
Subscribers: llvm-commits, wang0109, kubabrecka, chrisha
Differential Revision: https://reviews.llvm.org/D22339
llvm-svn: 275491
Summary:
These patterns are encounter when using instrumented DLL.
Without this patch, asan lit test are crashing when trying to hook
on RaiseException function.
Reviewers: rnk
Subscribers: llvm-commits, wang0109, chrisha
Differential Revision: https://reviews.llvm.org/D22340
llvm-svn: 275489
Summary:
The function wcslen is incorrectly hooked on windows 64-bits.
The interception library is not able to hook without breaking the code.
The function is too small and the interception must be done with
trampoline-hooking which turned out to be incorrect on a small
loop (first few instructions have a backedge).
Reviewers: rnk
Subscribers: wang0109, chrisha, llvm-commits, kubabrecka
Differential Revision: https://reviews.llvm.org/D22363
llvm-svn: 275488
We were quite happy to read past the end of the valid section data when
disassembling. Instead we entirely skip stub dylibs, and tell the user what's
happened if their section only has partial data.
llvm-svn: 275487
review it for consistency, accuracy, and clarity. These changes attempt to
address all of the above while keeping the text relatively terse.
<rdar://problem/24868841>
llvm-svn: 275485
Summary:
This aims to add support for __cpu_model and address Bug 25510. It uses
the code from lib/Support/Host.cpp for cpu detection, and creates
__cpu_model with that info.
Tested on OSX and built on Linux as well (though libgcc is the default).
The use of "asm" required -std=gnu99, hence the cmake change. Corrections
on better addressing this are welcome.
Previously reverted, up for review again to iron out outstanding issues.
Reviewers: llvm-commits, joerg, echristo, mehdi_amini
Subscribers: mehdi_amini
Differential Revision: http://reviews.llvm.org/D22181
llvm-svn: 275484
Fix for PR 28418.
opt never finishes compiling a test when -gvn option is passed.
The problem is caused by the fact that GVN fails to fold a constant expression.
Differential Revision: https://reviews.llvm.org/D22185
llvm-svn: 275483
trade-offs. When LLDB's multi-line editing support was first introduced
for expressions / REPL contexts the behavior was as follows:
* The Return key is treated as a line-break except at the end of the input
buffer, where a completeness test is applied
This worked well enough when writing code, and makes it trivial to insert
new lines above code you've already typed. Just use cursor navigation to
move up and type freely. Where it was awkward is that the gesture to insert
a line break and end editing is conflated for most people. Sometimes you
want Return to end the editing session and other times you want to insert
a line break.
This commit changes the behavior as follows:
* The Return key is treated as the end of editing except at the end of the
input buffer, where a completeness test is applied
* The Meta+Return sequence is always treated as a line break. This is
consistent with conventions in Facebook and elsewhere since
Alt/Option+Return is often mapped to Meta+Return. The unfortunate
exception is on macOS where this *can* be the case, but isn't by
default. Sigh.
Note that by design both before and after the patch pasting a Return
character always introduces a line break.
<rdar://problem/26886287>
llvm-svn: 275482
passed on the command line but never actually used. We consider a (top-level)
module to be used if any part of it is imported, either by the current
translation unit, or by any part of a top-level module that is itself used.
(Put another way, a module is used if an implicit modules build would have
loaded its .pcm file.)
llvm-svn: 275481
Summary:
When test was added we had no -fsanitize-address-use-after-scope in clang
so we had to use -mllvm -asan-use-after-scope=1.
Reviewers: eugenis
Subscribers: kubabrecka
Differential Revision: https://reviews.llvm.org/D22382
llvm-svn: 275475
Summary:
Invoke the weak/linkonce symbol resolution support (already used by
libLTO) that operates via the summary index.
This ensures prevailing linkonce are kept, by making them weak, and
marks preempted copies as available_externally when possible.
With this change, the older support for keeping the prevailing linkonce
(by changing their symbol resolution) is removed.
Reviewers: mehdi_amini
Subscribers: llvm-commits, mehdi_amini
Differential Revision: http://reviews.llvm.org/D22302
llvm-svn: 275474
We now collect all accesses with a constant stride, not just the ones with a
stride greater than one. This change was requested in the review of D19984.
llvm-svn: 275473
Changes to the underlying logging infrastructure in Fall 2016 Darwin
OSes were no longer showing up NSLog messages in command-line LLDB.
This change restores that functionality, and adds test cases to
verify the new behavior.
rdar://26732492
llvm-svn: 275472
This patch allows the formation of interleaved access groups in loops
containing predicated blocks. However, the predicated accesses are prevented
from forming groups.
Differential Revision: https://reviews.llvm.org/D19694
llvm-svn: 275471
For some reason it seems the second invocation is getting DMOD_OTHER_H
set to a path with/forward/slashes, but one of the use sites
has\back\slashes. There should be no difference with what was already
there, but for now try to avoid checking those paths.
llvm-svn: 275464
Summary:
When the expression is value dependent,
isIntegerConstantExpr() crashes in C++03 mode with
../tools/clang/lib/AST/ExprConstant.cpp:9330: (anonymous namespace)::ICEDiag CheckICE(const clang::Expr *, const clang::ASTContext &):
Assertion `!E->isValueDependent() && "Should not see value dependent exprs!"' failed.
In C++11 mode, that assert does not trigger.
This commit works around this in the check. We don't check
value-dependent indices and instead check their specialization.
Reviewers: alexfh, aaron.ballman
Subscribers: nemanjai, cfe-commits
Differential Revision: http://reviews.llvm.org/D22190
llvm-svn: 275461
distinct anonymous structs remain distinct despite having similar layout.
This is already ensured by distinguishing based on their placement in the parent
struct, using the function `findAnonymousStructOrUnionIndex`.
The problem is that this function only handles anonymous structs, like
```
class Foo { struct { int a; } }
```
and not untagged structs like
```
class Foo { struct { int a; } var; }
```
Both need to be handled, and this patch fixes that. The test case ensures that this functionality doesn't regress.
Thanks to Manman Ren for review.
https://reviews.llvm.org/D22270
llvm-svn: 275460
On Hexagon is it legal to packetize the instructions setting up call
arguments with the call instruction itself. This was already done,
except for tail calls. Make sure tail calls are handled as well.
llvm-svn: 275458
Peter Collingbourne