Summary:
The sanitizer symbolizers support printing the function offset
(difference between pc and function start) of a stackframe using the
`%q` format specifier.
Unfortunately this didn't actually work because neither the atos
or dladdr symbolizer set the `AddressInfo::function_offset` field.
This patch teaches both symbolizers to try to compute the function
offset. In the case of the atos symbolizer, atos might not report the
function offset (e.g. it reports a source location instead) so in this
case it fallsback to using `dladdr()` to compute the function offset.
Two test cases are included.
rdar://problem/56695185
Reviewers: kubamracek, yln
Subscribers: #sanitizers, llvm-commits
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D69549
The CUDA builtin library is apparently compiled in C++ mode, so the
assumption of convergent needs to be made in a typically non-SPMD
language. The functions in the library should still be assumed
convergent. Currently they are not, which is potentially incorrect and
this happens to work after the library is linked.
Summary:
Adds an optional parameter to `buildASTFromCodeWithArgs` that allows the user to
pass additional files that the main code needs to compile. This change makes
`buildASTFromCodeWithArgs` consistent with `runToolOnCodeWithArgs`.
Patch by Alexey Eremin.
Reviewers: gribozavr
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D70175
An advantage is that there are less portability concerns when writing
tests. For example, `-u` is not supported by all implementations of
`env`, but lit's internal shell provides its own `env` that supports
`-u`.
A disadvantage is that some shell constructs, such as parentheses, are
not supported, but FileCheck's test suite currently doesn't require
such constructs.
For comparison, lit configures its test suite in the same manner. See
`llvm/utils/lit/tests/lit.cfg`.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D70278
Based on D70020 by serge-sans-paille.
The ELF spec says:
> Furthermore, there may be internal references among these sections that would not make sense if one of the sections were removed or replaced by a duplicate from another object. Therefore, such groups must be included or omitted from the linked object as a unit. A section cannot be a member of more than one group.
GNU ld has 2 behaviors that we don't have:
- Group members (nextInSectionGroup != nullptr) are subject to garbage collection.
This includes non-SHF_ALLOC SHT_NOTE sections.
In particular, discarding non-SHF_ALLOC SHT_NOTE sections is an expected behavior by the Annobin
project. See
https://developers.redhat.com/blog/2018/02/20/annobin-storing-information-binaries/
for more information.
- Groups members are retained or discarded as a unit.
Members may have internal references that are not expressed as
SHF_LINK_ORDER, relocations, etc. It seems that we should be more conservative here:
if a section is marked live, mark all the other member within the
group.
Both behaviors are reasonable. This patch implements them.
A new field InputSectionBase::nextInSectionGroup tracks the next member
within a group. on ELF64, this increases sizeof(InputSectionBase) froms
144 to 152.
InputSectionBase::dependentSections tracks section dependencies, which
is used by both --gc-sections and /DISCARD/. We can't overload it for
the "next member" semantic, because we should allow /DISCARD/ to discard
sections independent of --gc-sections (GNU ld behavior). This behavior
may be reasonably used by `/DISCARD/ : { *(.ARM.exidx*) }` or `/DISCARD/
: { *(.note*) }` (new test `linkerscript/discard-group.s`).
Reviewed By: ruiu
Differential Revision: https://reviews.llvm.org/D70146
Cleanup handling of the denormal-fp-math attribute. Consolidate places
checking the allowed names in one place.
This is in preparation for introducing FP type specific variants of
the denormal-fp-mode attribute. AMDGPU will switch to using this in
place of the current hacky use of subtarget features for the denormal
mode.
Introduce a new header for dealing with FP modes. The constrained
intrinsic classes define related enums that should also be moved into
this header for uses in other contexts.
The verifier could use a check to make sure the denorm-fp-mode
attribute is sane, but there currently isn't one.
Currently, DAGCombiner incorrectly asssumes non-IEEE behavior by
default in the one current user. Clang must be taught to start
emitting this attribute by default to avoid regressions when this is
switched to assume ieee behavior if the attribute isn't present.
Summary:
This is a new warning which fires when one stores a reference to the
initializer_list contents in a way which may outlive the
initializer_list which it came from. In llvm this warning is triggered
whenever someone uses the initializer_list ArrayRef constructor.
This is indeed a dangerous thing to do (I myself was bitten by that at
least once), but it is not more dangerous than calling other ArrayRef
constructors with temporary objects -- something which we are used to
and have accepted as a tradeoff for ArrayRef's efficiency.
Currently, this warnings generates so much output that it completely
obscures any actionable warnings, so this patch disables it.
Reviewers: rnk, aaron.ballman
Subscribers: mgorny, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70122
The const-correctness of match() was fixed in rL372764, which allows
uses of Regex objects to be const in cases they couldn't be before. This
patch tightens up the const-ness of Regex in various such cases.
Reviewers: thopre
Reviewed By: thopre
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D68155
This patch implements writing function descriptors and TOC base into
data section, and also add function descriptors(both csect and label)
and TOC base symbols to the symbol table.
Summary:
readability-redundant-member-init removes redundant / unnecessary member and base class initialization. Unfortunately for the specific case of a copy constructor's initialization of a base class, gcc at strict warning levels warns if "base class is not initialized in the copy constructor of a derived class".
This patch adds an option `IgnoreBaseInCopyConstructors` defaulting to 0 (thus maintaining current behavior by default) to skip the specific case of removal of redundant base class initialization in the copy constructor. Enabling this option enables the resulting code to continue to compile successfully under `gcc -Werror=extra`. New test cases `WithCopyConstructor1` and `WithCopyConstructor2` in clang-tools-extra/test/clang-tidy/readability-redundant-member-init.cpp show that it removes redundant members even from copy constructors.
Reviewers: malcolm.parsons, alexfh, hokein, aaron.ballman, lebedev.ri
Patch by: poelmanc
Subscribers: mgehre, lebedev.ri, cfe-commits
Tags: #clang, #clang-tools-extra
Differential revision: https://reviews.llvm.org/D69145
As discussed in D70148 (and caused a revert of the original commit):
if we insert at the select, then we can produce invalid IR because
the replacement for the compare may have uses before the select.
This fills in the small family of MVE intrinsics that have nothing to
do with vectors: they implement bit-shift operations on 32- or 64-bit
values held in one or two general-purpose registers. Most of these
shift operations saturate if shifting left, and round to nearest if
shifting right, although LSLL and ASRL behave like ordinary shifts.
When these instructions take a variable shift count in a register,
they pay attention to its sign, so that (for example) LSLL or UQRSHLL
will shift left if given a positive number but right if given a
negative one. That makes even LSLL and ASRL different enough from
standard LLVM IR shift semantics that I couldn't see any better
alternative than to simply model the whole family as a set of
MVE-specific IR intrinsics.
(The //immediate// forms of LSLL and ASRL, on the other hand, do
behave exactly like a standard IR shift of a 64-bit value. In fact,
those forms don't have ACLE intrinsics defined at all, because you can
just write an ordinary C shift operation if you want one of those.)
The 64-bit shifts have to be instruction-selected in C++, because they
deliver two output values. But the 32-bit ones are simple enough that
I could write a DAG isel pattern directly into each Instruction
record.
Reviewers: ostannard, MarkMurrayARM, dmgreen
Reviewed By: dmgreen
Subscribers: kristof.beyls, hiraditya, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D70319
Start moving towards treating this as a property of the calling
convention, and not the subtarget. The default denormal mode should
not be part of the subtarget, and be moved into a separate function
attribute.
This patch is still NFC. The denormal mode remains as a subtarget
feature for now, but make the necessary changes to switch to using an
attribute.
Summary:
This is mostly mechanical, with a few exceptions:
- getDeducedType moved into AST.h where it belongs. It now takes
ASTContext instead of ParsedAST, and avoids using the preprocessor.
- hover now uses SelectionTree directly rather than via
getDeclAtPosition helper
- hover on 'auto' used to find the decl that contained the 'auto' and
use that to set Kind and documentation for the hover result.
Now we use targetDecl() to find the decl matching the deduced type instead.
This changes tests, e.g. 'variable' -> class for auto on lambdas.
I think this is better, but the motivation was to avoid depending on
the internals of DeducedTypeVisitor. This functionality is removed
from the visitor.
Reviewers: kadircet
Subscribers: mgorny, ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D70357
AMDGPU needs to know the FP mode for the function to answer this
correctly when this is removed from the subtarget.
AArch64 had to make this more complicated by using this from an IR
hook, so add an IR typed overload.
Summary:
Most of IR instructions got better code size estimations after commit 47a5c36b.
So default parameters values should be updated to improve inlining and
unrolling for the target.
Reviewers: rampitec, arsenm
Reviewed By: rampitec
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, zzheng, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70391
I just used the mangled names as this test is anyway a Darwin-only ObjC++ test.
We probably should also test this on other platforms but that will be
another commit as we need to untangle the ObjC and C++ parts first.
Summary:
In addition to adding `override` wherever possible, clang-tidy's `modernize-use-override` nicely removes `virtual` when `override` or `final` is specified, and further removes override when final is specified. While this is great default behavior, when code needs to be compiled with gcc at high warning levels that include `gcc -Wsuggest-override` or `gcc -Werror=suggest-override`, clang-tidy's removal of the redundant `override` keyword causes gcc to emit a warning or error. This discrepancy / conflict has been noted by others including a comment on Stack Overflow and by Mozilla's Firefox developers.
This patch adds an AllowOverrideAndFinal option defaulting to 0 - thus preserving current behavior - that when enabled allows both `override` and `final` to co-exist, while still fixing all other issues.
The patch includes a test file verifying all combinations of virtual/override/final, and mentions the new option in the release notes.
Reviewers: alexfh, djasper, JonasToth
Patch by: poelmanc
Subscribers: JonasToth, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D70165
Summary:
`readability-redundant-string-init` was one of several clang-tidy checks documented as failing for C++17. (The failure mode in C++17 is that it changes `std::string Name = ""`; to `std::string Name = Name;`, which actually compiles but crashes at run-time.)
Analyzing the AST with `clang -Xclang -ast-dump` showed that the outer `CXXConstructExprs` that previously held the correct SourceRange were being elided in C++17/2a, but the containing `VarDecl` expressions still had all the relevant information. So this patch changes the fix to get its source ranges from `VarDecl`.
It adds one test `std::string g = "u", h = "", i = "uuu", j = "", k;` to confirm proper warnings and fixit replacements in a single `DeclStmt` where some strings require replacement and others don't. The readability-redundant-string-init.cpp and readability-redundant-string-init-msvc.cpp tests now pass for C++11/14/17/2a.
Reviewers: gribozavr, etienneb, alexfh, hokein, aaron.ballman, gribozavr2
Patch by: poelmanc
Subscribers: NoQ, MyDeveloperDay, Eugene.Zelenko, dylanmckay, cfe-commits
Tags: #clang, #clang-tools-extra
Differential Revision: https://reviews.llvm.org/D69238
Summary: Matches GNU objdump. Makes debugging easier for me as I'm working out addresses from symbol+addend, so it would be good to be calculating in a single format.
Reviewers: MaskRay, grimar, jhenderson, bd1976llvm
Reviewed By: jhenderson
Subscribers: sdardis, jrtc27, atanasyan, rupprecht, seiya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D69997
As detailed in PR43971/D70267, the use of XFormVExtractWithShuffleIntoLoad causes issues where we end up in infinite loops of extract(targetshuffle(vecload)) -> extract(shuffle(vecload)) -> extract(vecload) -> extract(targetshuffle(vecload)), there are just too many legalization checks at every stage that we can't guarantee that extract(shuffle(vecload)) -> scalarload can occur.
At the moment we see a number of minor regressions as we don't fold extract(shuffle(vecload)) -> scalarload before legal ops, these can be addressed in future patches and extension of X86ISelLowering's combineExtractWithShuffle.
Summary:
This is some really shady code. It's supposed to kick in after an expression already failed and then try to look
up "unknown types" that for some undocumented reason can't be resolved during/before parsing. Beside the
fact that we never mark any type as `EVUnknownType` in either swift-lldb or lldb (which means this code is unreachable),
this code doesn't even make the expression evaluation succeed if if would ever be executed but instead seems
to try to load more debug info that maybe any following expression evaluations might succeed.
This patch removes ClangExpressionDeclMap::ResolveUnknownTypes and the related data structures/checks/calls.
Reviewers: davide
Reviewed By: davide
Subscribers: aprantl, abidh, JDevlieghere, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D70388
Summary:
Assert in getFunctionLocalOffsetAfterInsn() fails when processing a call
MachineInstr inside a bundle and compiling with debug info. This is
because labels are added by DwarfDebug::beginInstruction() which is
called for each top-level MI by EmitFunctionBody()'s for-loop iteration
but constructCallSiteEntryDIEs() which calls
getFunctionLocalOffsetAfterInsn() iterates over all MIs.
This commit modifies constructCallSiteEntryDIEs() to get the associated
bundle MI for call MIs inside a bundle and use that to when calling
getFunctionLocalOffsetAfterInsn() and getLabelAfterInsn(). It also skips
loop iterations for bundle MIs since the loop statements are concerned
with debug info for each physical instructions and bundles represent a
group of instructions. It also fix the comment about PCAddr since the
code is getting the return address and not the call address.
Reviewers: dstenb, vsk, aprantl, djtodoro, dblaikie, NikolaPrica
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70293
Jonas Devlieghere