This matches how GCC handles it, see e.g. https://gcc.godbolt.org/z/HPplnl.
GCC documents the gnu_inline attribute with "In C++, this attribute does
not depend on extern in any way, but it still requires the inline keyword
to enable its special behavior."
The previous behaviour of gnu_inline in C++, without the extern
keyword, can be traced back to the original commit that added
support for gnu_inline, SVN r69045.
Differential Revision: https://reviews.llvm.org/D67414
llvm-svn: 373078
Instead of returning an optional, just return the input string if
demangling fails, as that's what all callers use anyway.
Differential Revision: https://reviews.llvm.org/D68015
llvm-svn: 373077
This avoids a few lines of boilerplate of dealing with C string
allocations.
Add a testcase for a case where demangling shouldn't happen.
Differential Revision: https://reviews.llvm.org/D68014
llvm-svn: 373076
ModuleList.cpp includes clang/Driver/Driver.h. Reflect that in the build
system. Not having this can cause build failures if ModuleList.cpp is
built before Driver.inc is generated.
llvm-svn: 373073
Summary:
Windows unwinding is weird. The unwind rules do not (always) describe
the precise layout of the stack, but rather expect the debugger to scan
the stack for something which looks like a plausible return address, and
the unwind based on that. The reason this works somewhat reliably is
because the the unwinder also has access to the frame sizes of the
functions on the stack. This allows it (in most cases) to skip function
pointers in local variables or function arguments, which could otherwise
be mistaken for return addresses.
Implementing this kind of unwind mechanism in lldb was a bit challenging
because we expect to be able to statically describe (in the UnwindPlan)
structure, the layout of the stack for any given instruction. Giving a
precise desription of this is not possible, because it requires
correlating information from two functions -- the pushed arguments to a
function are considered a part of the callers stack frame, and their
size needs to be considered when unwinding the caller, but they are only
present in the unwind entry of the callee. The callee may end up being
in a completely different module, or it may not even be possible to
determine it statically (indirect calls).
This patch implements this functionality by introducing a couple of new
APIs:
SymbolFile::GetParameterStackSize - return the amount of stack space
taken up by parameters of this function.
SymbolFile::GetOwnFrameSize - the size of this function's frame. This
excludes the parameters, but includes stuff like local variables and
spilled registers.
These functions are then used by the unwinder to compute the estimated
location of the return address. This address is not always exact,
because the stack may contain some additional values -- for instance, if
we're getting ready to call a function then the stack will also contain
partially set up arguments, but we will not know their size because we
haven't called the function yet. For this reason the unwinder will crawl
up the stack from the return address position, and look for something
that looks like a possible return address. Currently, we assume that
something is a valid return address if it ends up pointing to an
executable section.
All of this logic kicks in when the UnwindPlan sets the value of CFA as
"isHeuristicallyDetected", which is also the final new API here. Right
now, only SymbolFileBreakpad implements these APIs, but in the future
SymbolFilePDB will use them too.
Differential Revision: https://reviews.llvm.org/D66638
llvm-svn: 373072
Previously, these were unseen because the wrapper script would swallow
them. This fixes the following types of warnings:
- methods being declared more than once
- swig complained about ignoring operator=, so I just removed it
llvm-svn: 373069
Summary:
Creates a new darwin ClangTidy module and adds the darwin-dispatch-once-nonstatic check that warns about dispatch_once_t variables not in static or global storage. This catches a missing static for local variables in e.g. singleton initialization behavior, and also warns on storing dispatch_once_t values in Objective-C instance variables. C/C++ struct/class instances may potentially live in static/global storage, and are ignored for this check.
The osx.API static analysis checker can find the non-static storage use of dispatch_once_t; I thought it useful to also catch this issue in clang-tidy when possible.
This is a re-land of https://reviews.llvm.org/D67567
Reviewers: thakis, gribozavr, stephanemoore
Subscribers: Eugene.Zelenko, mgorny, xazax.hun, jkorous, arphaman, kadircet, usaxena95
Tags: #clang-tools-extra, #clang, #llvm
Differential Revision: https://reviews.llvm.org/D68109
llvm-svn: 373065
D64572 / rL365818 changed the way that the file paths were collected, which meant we lost the file pattern expansion necessary when working with DOS command prompt
llvm-svn: 373062
This caused severe compile-time regressions, see PR43455.
> Modern processors predict the targets of an indirect branch regardless of
> the size of any jump table used to glean its target address. Moreover,
> branch predictors typically use resources limited by the number of actual
> targets that occur at run time.
>
> This patch changes the semantics of the option `-max-jump-table-size` to limit
> the number of different targets instead of the number of entries in a jump
> table. Thus, it is now renamed to `-max-jump-table-targets`.
>
> Before, when `-max-jump-table-size` was specified, it could happen that
> cluster jump tables could have targets used repeatedly, but each one was
> counted and typically resulted in tables with the same number of entries.
> With this patch, when specifying `-max-jump-table-targets`, tables may have
> different lengths, since the number of unique targets is counted towards the
> limit, but the number of unique targets in tables is the same, but for the
> last one containing the balance of targets.
>
> Differential revision: https://reviews.llvm.org/D60295
llvm-svn: 373060
Summary:
The const-correctness of match() was fixed in rL372764, which allows
such static Regex objects to be marked const.
Reviewers: thopre
Reviewed By: thopre
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68091
llvm-svn: 373058
The following code
```
struct f {
template <class T>
void bar() && noexcept {}
};
```
will be formatted to the following with LLVM style, and
`AlwaysBreakTemplateDeclarations: Yes`
```
struct f {
template <class T>
void bar() && noexcept {}
};
```
The indentation of the `void bar()` line is wrong.
Patch by Andreas Wass (AndWass)!
Differential Revision: https://reviews.llvm.org/D68072
llvm-svn: 373056
A lot of commands are made up of CommandObjectMultiword with
subcommands. CommandObjectMultiword actually has some functionality
on its own that wasn't tested before.
llvm-svn: 373050
Summary:
Right now latency generation can incorrectly select the scratch register
as a dependency-carrying register.
- Move the logic for preventing register selection from Uops
implementation to common SnippetGenerator class.
- Aliasing detection now takes a set of forbidden registers just like
random register assignment does.
Reviewers: gchatelet
Subscribers: tschuett, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68084
llvm-svn: 373048
hasDedicatedExits.
For the compile time problem described in https://reviews.llvm.org/D67359,
turns out the root cause is there are many duplicates in ExitBlocks so
the algorithm complexity of hasDedicatedExits gets very high. If we remove
the duplicates, the compile time issue is gone.
Thanks to Philip Reames for raising a good question and it leads me to
find the root cause.
Differential Revision: https://reviews.llvm.org/D68107
llvm-svn: 373045
exits"
Get a better approach in https://reviews.llvm.org/D68107 to solve the problem.
Revert the initial patch and will commit the new one soon.
This reverts commit rL372990.
llvm-svn: 373044
We need to discard all remaining cleanups if an earlier cleanup failed,
otherwise we may try to rerun the remaining cleanups later, potentially
after the scope containing the object is destroyed. (This can happen
when checking a potential constant expression.)
llvm-svn: 373042
We can't use short granules with stack instrumentation when targeting older
API levels because the rest of the system won't understand the short granule
tags stored in shadow memory.
Moreover, we need to be able to let old binaries (which won't understand
short granule tags) run on a new system that supports short granule
tags. Such binaries will call the __hwasan_tag_mismatch function when their
outlined checks fail. We can compensate for the binary's lack of support
for short granules by implementing the short granule part of the check in
the __hwasan_tag_mismatch function. Unfortunately we can't do anything about
inline checks, but I don't believe that we can generate these by default on
aarch64, nor did we do so when the ABI was fixed.
A new function, __hwasan_tag_mismatch_v2, is introduced that lets code
targeting the new runtime avoid redoing the short granule check. Because tag
mismatches are rare this isn't important from a performance perspective; the
main benefit is that it introduces a symbol dependency that prevents binaries
targeting the new runtime from running on older (i.e. incompatible) runtimes.
Differential Revision: https://reviews.llvm.org/D68059
llvm-svn: 373035
Haojian Wu