This patch fixes a bug where exceptions in 32 bit builds
would be incorrectly aligned because malloc only provides 8 byte aligned
memory where 16 byte alignment is needed.
This patch makes libc++abi correctly use posix_memalign when it's
available. This requires defining _LIBCPP_BUILDING_LIBRARY so that
libc++ only defines _LIBCPP_HAS_NO_ALIGNED_ALLOCATION when libc doesn't
support it and not when aligned new/delete are disable for other
reasons.
This bug somehow made it into the 7.0 release, making it a regression.
Therefore this patch should be included in the next dot release.
llvm-svn: 342815
Summary:
r342805 added support for the check-cxx-abilist target on FreeBSD, but broke
the target on macOS in doing so. The problem is that the GENERIC_TARGET_TRIPLE
gets overwritten after replacing the FreeBSD regular expression, which
nullifies the replacement done with the darwin regular expression.
Reviewers: dim, EricWF
Subscribers: emaste, mgorny, krytarowski, christof, dexonsmith, cfe-commits, libcxx-commits
Differential Revision: https://reviews.llvm.org/D52394
llvm-svn: 342813
aarch64 testing is broken because "medium" is not a valid
code-model on aarch64, and codemodels.c tests that. This fixes
that problem by adding "-triple x86_64-unknown-linux-gnu" to the
test with "-mcode-model moedium".
llvm-svn: 342812
Summary:
They are introduced in r338479; their Linux ABI changes are recorded in r338486.
TODO: Record the Mac OS X ABI changes.
Reviewers: EricWF
Reviewed By: EricWF
Subscribers: christof, ldionne, libcxx-commits, cfe-commits
Differential Revision: https://reviews.llvm.org/D52391
llvm-svn: 342810
Summary:
The `[[nodiscard]]` attribute is intended to help users find bugs where
function return values are ignored when they shouldn't be. After C++17 the
C++ standard has started to declared such library functions as `[[nodiscard]]`.
However, this application is limited and applies only to dialects after C++17.
Users who want help diagnosing misuses of STL functions may desire a more
liberal application of `[[nodiscard]]`.
For this reason libc++ provides an extension that does just that! The
extension must be enabled by defining `_LIBCPP_ENABLE_NODISCARD`. The extended
applications of `[[nodiscard]]` takes two forms:
1. Backporting `[[nodiscard]]` to entities declared as such by the
standard in newer dialects, but not in the present one.
2. Extended applications of `[[nodiscard]]`, at the libraries discretion,
applied to entities never declared as such by the standard.
Users may also opt-out of additional applications `[[nodiscard]]` using
additional macros.
Applications of the first form, which backport `[[nodiscard]]` from a newer
dialect may be disabled using macros specific to the dialect it was added. For
example `_LIBCPP_DISABLE_NODISCARD_AFTER_CXX17`.
Applications of the second form, which are pure extensions, may be disabled
by defining `_LIBCPP_DISABLE_NODISCARD_EXT`.
This patch was originally written by me (Roman Lebedev),
then but then reworked by Eric Fiselier.
Reviewers: mclow.lists, thakis, EricWF
Reviewed By: thakis, EricWF
Subscribers: llvm-commits, mclow.lists, lebedev.ri, EricWF, rjmccall, Quuxplusone, cfe-commits, christof
Differential Revision: https://reviews.llvm.org/D45179
llvm-svn: 342808
Follow-up to rL342324 (D52059):
Missing optimizations with blendv are shown in:
https://bugs.llvm.org/show_bug.cgi?id=38814
This is an easier and more powerful solution than adding pattern matching for a few
special cases in the backend. The potential danger with this transform in IR is that
the condition value can get separated from the select, and the backend might not be
able to make a blendv out of it again.
llvm-svn: 342806
numbers off of freebsd target triples, when generating the name of the
ABI list file for check-cxx-abilist target.
Also remove unnecessary parentheses in the regex for darwin, and
slightly reword the comment.
llvm-svn: 342805
AbsPosToLineColumnPos is the only reader of m_user_expression_start_pos
and actually treats it like a size_t. Also the value we store in
m_user_expression_start_pos is originally a size_t, so it makes sense
to change the type of this variable to size_t.
llvm-svn: 342804
DWARF5 spec says about single file split case:
"The sections that do not require relocation, however, can be written
to the relocatable object (.o) file but ignored by the
the linker or they can be written to a separate DWARF object (.dwo) file
that need not be accessed by the linker."
Nice way to make linker to ignore them is to set SHF_EXCLUDE flag.
It seems to be not harmful to always set it for .dwo sections.
That is what this patch does.
Differential revision: https://reviews.llvm.org/D52303
llvm-svn: 342800
Summary: This restores the combine that was reverted in r341883. The infinite loop from the failing test no longer occurs due to changes from r342163.
Reviewers: spatel, dmgreen
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D52070
llvm-svn: 342797
Summary: Similar to D51893 which was for memcpy
Reviewers: efriedma
Reviewed By: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D52063
llvm-svn: 342796
We don't have a vXi8 shift left so we need to bitcast to a vXi16 vector to perform the shift. If we let lowering legalize the vXi8 shift we get an extra and that we don't need and fail to remove.
llvm-svn: 342795
Objects are determined to be smart pointers if they have both a star and arrow
operator. Some implementations of smart pointers have these overloaded
operators in a base class, while the check only searched the derived class.
This fix will also look for the operators in the base class.
llvm-svn: 342794
This can be used to detect whether the code is being built with UBSan using
the __has_feature(undefined_behavior_sanitizer) predicate.
Differential Revision: https://reviews.llvm.org/D52386
llvm-svn: 342793
Not used productively, so no observable functional change.
Note that printSCFG doesn't yet work reliably, it seems to crash
sometimes.
llvm-svn: 342790
A recent commit I made broke aarch64 testing, because "kernel"
apparently is not a valid code-model on aarch64, and one of my tests
tested that. This fixes the problem (hopefully) by adding "-triple
x86_64-unknown-linux-gnu" to the test build with "-mcodel-model
kernel".
Differential Revision: https://reviews.llvm.org/D52383
llvm-svn: 342789
Previously we used SUBREG_TO_REG+MOV32ri. But regular isel was changed recently to use the MOV32ri64 pseudo. Fast isel now does the same.
llvm-svn: 342788
Support for vectorizing loops with secondary floating-point induction
variables was added in r276554. A primary integer IV is still required
for vectorization to be done. If an FP IV was found, but no integer IV
was found at all (primary or secondary), the attempt to vectorize still
went forward, causing a compiler-crash. This change abandons that
attempt when no integer IV is found. (Vectorizing FP-only cases like
this, rather than bailing out, is discussed as possible future work
in D52327.)
See PR38800 for more information.
Differential Revision: https://reviews.llvm.org/D52327
llvm-svn: 342786
has been finalized.
This prevents crashes on unfinalized objects for clients using
JITEventListeners.
Patch by Geoff Levner. Thanks Geoff!
llvm-svn: 342783
This allows the native reader to find records of class/struct/
union type and dump them. This behavior is tested by using the
diadump subcommand against golden output produced by actual DIA
SDK on the same PDB file, and again using pretty -native to
confirm that we actually dump the classes. We don't find class
members or anything like that yet, for now it's just the class
itself.
llvm-svn: 342779
As a prerequisite to time-passes implementation which needs to time both passes
and analyses, adding instrumentation points to the Analysis Manager.
The are two functional differences between Pass and Analysis instrumentation:
- the latter does not increment pass execution counter
- it does not provide ability to skip execution of the corresponding analysis
Reviewers: chandlerc, philip.pfaffe
Differential Revision: https://reviews.llvm.org/D51275
llvm-svn: 342778
GNU binutils import libraries aren't the same kind of short import
libraries as link.exe and LLD produce, but are a plain static library
containing .idata section chunks. MSVC link.exe can successfully link
to them.
In order for imports from GNU import libraries to mix properly with the
normal import chunks, the chunks from the existing mechanism needs to
be added into named sections like .idata$2.
These GNU import libraries consist of one header object, a number of
object files, one for each imported function/variable, and one trailer.
Within the import libraries, the object files are ordered alphabetically
in this order. The chunks stemming from these libraries have to be
grouped by what library they originate from and sorted, to make sure
the section chunks for headers and trailers for the lists are ordered
as intended. This is done on all sections named .idata$*, before adding
the synthesized chunks to them.
Differential Revision: https://reviews.llvm.org/D38513
llvm-svn: 342777
Instead of indexing local variables by DIE offset, use the variable
name + the path through the lexical block tree. This makes the lookup
key consistent across duplicate abstract origins in different CUs.
llvm-svn: 342776
Comparison functions used in sorting algorithms need to have strict weak
ordering. Remove the assert and allow comparisons on all lists.
llvm-svn: 342774
Summary:
By using the existing isCodeGenOnly bit in the tablegen defs, as
suggested by tlively in https://reviews.llvm.org/D51662
Tested: llvm-lit -v `find test -name WebAssembly`
Reviewers: tlively
Subscribers: dschuff, sbc100, jgravelle-google, aheejin, sunfish, llvm-commits
Differential Revision: https://reviews.llvm.org/D52373
llvm-svn: 342772
Modify the RetainCountChecker to perform state "adjustments" in
checkEndFunction, as performing work in PreStmt<ReturnStmt> does not
work with destructors.
The previous version made an implicit assumption that no code runs
after the return statement is executed.
rdar://43945028
Differential Revision: https://reviews.llvm.org/D52338
llvm-svn: 342770
If the non-sink report is generated at the exit node, it will be
suppressed by the current functionality in isInevitablySinking, as it
only checks the successors of the block, but not the block itself.
The bug shows up in RetainCountChecker checks.
Differential Revision: https://reviews.llvm.org/D52284
llvm-svn: 342766