Justin Lebar [Thu, 29 Dec 2016 19:59:26 +0000 (19:59 +0000)]
[ADT] Delete RefCountedBaseVPTR.
Summary:
This class is unnecessary.
Its comment indicated that it was a compile error to allocate an
instance of a class that inherits from RefCountedBaseVPTR on the stack.
This may have been true at one point, but it's not today.
Moreover you really do not want to allocate *any* refcounted object on
the stack, vptrs or not, so if we did have a way to prevent these
objects from being stack-allocated, we'd want to apply it to regular
RefCountedBase too, obviating the need for a separate RefCountedBaseVPTR
class.
It seems that the main way RefCountedBaseVPTR provides safety is by
making its subclass's destructor virtual. This may have been helpful at
one point, but these days clang will emit an error if you define a class
with virtual functions that inherits from RefCountedBase but doesn't
have a virtual destructor.
Teresa Johnson [Wed, 28 Dec 2016 18:00:08 +0000 (18:00 +0000)]
[ThinLTO] No need to rediscover imports in distributed backend
Summary:
We can simply import all external values with summaries included in
the individual index file created for the distributed backend job,
as only those are added to the individual index file created by the
WriteIndexesThinBackend (in addition to summaries for the original
module, which are skipped here).
While computing the cross module imports on this index would come to
the same conclusion as the original thin link import logic, it is
unnecessary work. And when tuning, it avoids the need to pass the
same function importing parameters (e.g. -import-instr-limit) to
both the thin link and the backends (otherwise they won't make the
same decisions).
Our newly aggressive constant folding logic makes it possible for
CGExprConstant to see the same CompoundLiteralExpr more than once. So,
emitting a new GlobalVariable every time we see a CompoundLiteral is no
longer correct.
We had a similar issue with BlockExprs that was caught while testing
said aggressive folding, so I applied the same style of fix (see D26410)
here. If we find yet another case where this needs to happen, we should
probably refactor this so we don't have a third DenseMap+getter+setter.
As a design note: getAddrOfConstantCompoundLiteralIfEmitted is really
only intended to be called by ConstExprEmitter::EmitLValue. So,
returning a GlobalVariable* instead of a ConstantAddress costs us
effectively nothing, and saves us either a few bytes per entry in our
map or a bit of code duplication.
Richard Smith [Wed, 28 Dec 2016 06:27:18 +0000 (06:27 +0000)]
Mark 'auto' as dependent when instantiating the type of a non-type template
parameter. Fixes failed deduction for 'auto' non-type template parameters
nested within templates.
Richard Smith [Wed, 28 Dec 2016 02:37:25 +0000 (02:37 +0000)]
DR1315: a non-type template argument in a partial specialization is permitted
to make reference to template parameters. This is only a partial
implementation; we retain the restriction that the argument must not be
type-dependent, since it's unclear how that would work given the existence of
other language rules requiring an exact type match in this context, even for
type-dependent cases (a question has been raised on the core reflector).
David Blaikie [Tue, 27 Dec 2016 22:05:35 +0000 (22:05 +0000)]
DebugInfo: Don't include size/alignment on class declarations
This seems like it must've been a leftover by accident - no tests were
backing it up & it doesn't make much sense to include size/alignment on
class declarations (it'd only be on those declarations for which the
definition was available - otherwise the size/alignment would not be
known).
Richard Smith [Tue, 27 Dec 2016 20:03:09 +0000 (20:03 +0000)]
Add warning flag for "partial specialization is not more specialized than primary template" error (since Eigen hits it), and while I'm here also add a warning flag for "partial specialization is not usable because one or more of its parameters cannot be deduced" warning.
[DOXYGEN] Improved doxygen comments for xmmintrin.h intrinsics.
Added \n commands to insert a line breaks where necessary, since one long line of documentation is nearly unreadable.
Formatted comments to fit into 80 chars.
In some cases added \a command in front of the parameter names to display them in italics.
Richard Smith [Tue, 27 Dec 2016 07:56:27 +0000 (07:56 +0000)]
DR1495: A partial specialization is ill-formed if it is not (strictly) more
specialized than the primary template. (Put another way, if we imagine there
were a partial specialization matching the primary template, we should never
select it if some other partial specialization also matches.)
Richard Smith [Tue, 27 Dec 2016 06:14:37 +0000 (06:14 +0000)]
Work around a standard defect: template argument deduction for non-type
template parameters of reference type basically doesn't work, because we're
always deducing from an argument expression of non-reference type, so the type
of the deduced expression never matches. Instead, compare the type of an
expression naming the parameter to the type of the argument.
Windows uses PE/COFF which is inherently position independent. The use
of the PIC model is unnecessary. In fact, we would generate invalid
code using the ELF PIC model when PIC was enabled previously. Now that
we no longer accept -fPIC and -fpoc, this switches the internal
representation to the static model to permit us to make PIC modules
invalid when targeting Windows. This should not change the code
generation, only the internal state management.
Richard Smith [Tue, 27 Dec 2016 02:02:09 +0000 (02:02 +0000)]
Check and build conversion sequences for non-type template arguments in
dependent contexts when processing the template in C++11 and C++14, just like
we do in C++98 and C++1z. This allows us to diagnose invalid templates earlier.
[DOXYGEN] Improved doxygen comments for x86 intrinsics.
Improved doxygen comments for the following intrinsics headers: __wmmintrin_pclmul.h, bmiintrin.h, emmintrin.h, f16cintrin.h, immintrin.h, mmintrin.h, pmmintrin.h, tmmintrin.h
Added \n commands to insert a line breaks where necessary, since one long line of documentation is nearly unreadable.
Formatted comments to fit into 80 chars.
In some cases added \a command in front of the parameter names to display them in italics.
Chandler Carruth [Tue, 27 Dec 2016 00:31:34 +0000 (00:31 +0000)]
[PM] The new pass manager requires a registered target for these, and
given that they hard code specific triples that seems reasonable so add
the REQUIRES.
Chandler Carruth [Tue, 27 Dec 2016 00:13:09 +0000 (00:13 +0000)]
[PH] Teach the new PM code path to support -disable-llvm-passes.
This is kind of funny because I specifically did work to make this easy
and then it didn't actually get implemented.
I've also ported a set of tests that rely on this functionality to run
with the new PM as well as the old PM so that we don't mess this up in
the future.
Marina Yatsina [Mon, 26 Dec 2016 12:23:42 +0000 (12:23 +0000)]
[inline-asm]No error for conflict between inputs\outputs and clobber list
According to extended asm syntax, a case where the clobber list includes a variable from the inputs or outputs should be an error - conflict.
for example:
Driver: warn on -fPIC/-fpic/-fPIE/-fpie on Windows
Use of these flags would result in the use of ELF-style PIE/PIC code
which is incorrect on Windows. Windows is inherently PIC by means of
the DLL slide that occurs at load. This also mirrors the behaviour on
GCC for MinGW. Currently, the Windows x86_64 forces the relocation
model to PIC (Level 2). This is unchanged for now, though we should
remove any assumptions on that and change it to a static relocation
model.
Richard Smith [Sun, 25 Dec 2016 08:05:23 +0000 (08:05 +0000)]
Fix some subtle wrong partial ordering bugs particularly with C++1z auto-typed
non-type template parameters.
During partial ordering, when checking the substituted deduced template
arguments match the original, check the types of non-type template arguments
match even if they're dependent. The only way we get dependent types here is if
they really represent types of the other template (which are supposed to be
modeled as being substituted for unique, non-dependent types).
In order to make this work for auto-typed non-type template arguments, we need
to be able to perform auto deduction even when the initializer and
(potentially) the auto type are dependent, support for which is the bulk of
this patch. (Note that this requires the ability to deduce only a single level
of a multi-level dependent type.)
Anton Yartsev [Sun, 25 Dec 2016 00:57:51 +0000 (00:57 +0000)]
Fix for PR15623 (corrected r290413 reverted at 290415). The patch eliminates unwanted ProgramState checker data propagation from an operand of the logical operation to operation result.
The patch also simplifies an assume of a constraint of the form: "(exp comparison_op expr) != 0" to true into an assume of "exp comparison_op expr" to true. (And similarly, an assume of the form "(exp comparison_op expr) == 0" to true as an assume of exp comparison_op expr to false.) which improves precision overall.
https://reviews.llvm.org/D22862
Malcolm Parsons [Sat, 24 Dec 2016 13:35:14 +0000 (13:35 +0000)]
[ASTMatchers] Add hasInClassInitializer traversal matcher for FieldDecl.
Summary:
I needed to know whether a FieldDecl had an in-class
initializer for D26453. I used a narrowing matcher there, but a
traversal matcher might be generally useful.
Richard Smith [Sat, 24 Dec 2016 04:09:05 +0000 (04:09 +0000)]
When producing a name of a partial specialization in a diagnostic, use the
template arguments as written rather than the canonical template arguments,
so we print more user-friendly names for template parameters.
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics.
Tagged parameter names with \a doxygen command to display parameters in italics.
Added \n commands to insert a line break to make the documentation more readable.
Formatted comments to fit into 80 chars.
Chandler Carruth [Fri, 23 Dec 2016 20:44:01 +0000 (20:44 +0000)]
[PM] Introduce options to enable the (still experimental) new pass
manager, and a code path to use it.
The option is actually a top-level option but does contain
'experimental' in the name. This is the compromise suggested by Richard
in discussions. We expect this option will be around long enough and
have enough users towards the end that it merits not being relegated to
CC1, but it still needs to be clear that this option will go away at
some point.
The backend code is a fresh codepath dedicated to handling the flow with
the new pass manager. This was also Richard's suggested code structuring
to essentially leave a clean path for development rather than carrying
complexity or idiosyncracies of how we do things just to share code with
the parts of this in common with the legacy pass manager. And it turns
out, not much is really in common even though we use the legacy pass
manager for codegen at this point.
I've switched a couple of tests to run with the new pass manager, and
they appear to work. There are still plenty of bugs that need squashing
(just with basic experiments I've found two already!) but they aren't in
this code, and the whole point is to expose the necessary hooks to start
experimenting with the pass manager in more realistic scenarios.
That said, I want to *strongly caution* anyone itching to play with
this: it is still *very shaky*. Several large components have not yet
been shaken down. For example I have bugs in both the always inliner and
inliner that I have already spotted and will be fixing independently.
Still, this is a fun milestone. =D
One thing not in this patch (but that might be very reasonable to add)
is some level of support for raw textual pass pipelines such as what
Sean had a patch for some time ago. I'm mostly interested in the more
traditional flow of getting the IR out of Clang and then running it
through opt, but I can see other use cases so someone may want to add
it.
And of course, *many* features are not yet supported!
- O1 is currently more like O2
- None of the sanitizers are wired up
- ObjC ARC optimizer isn't wired up
- ...
Egor Churaev [Fri, 23 Dec 2016 16:11:25 +0000 (16:11 +0000)]
[OpenCL] Align fake address space map with the SPIR target maps.
Summary:
We compile user opencl kernel code with spir triple. But built-ins are written in OpenCL and we compile it with triple x86_64 to be able to use x86 intrinsics. And we need address spaces to match in both cases. So, we change fake address space map in OpenCL for matching with spir.
On CPU address spaces are not really important but we'd like to preserve address space information in order to perform optimizations relying on this info like enhanced alias analysis.
Chandler Carruth [Fri, 23 Dec 2016 05:19:47 +0000 (05:19 +0000)]
Add an assert to catch improperly constructed %diff sequences in
diagnostics and fix one such diagnostic.
Sadly, this assert doesn't catch this bug because we have no tests that
emit this diagnostic! Doh! I'm following up on the commit that
introduces it to get that fixed. Then this assert will help in a more
direct way.
Anton Yartsev [Fri, 23 Dec 2016 03:31:00 +0000 (03:31 +0000)]
Fix for PR15623. The patch eliminates unwanted ProgramState checker data propagation from an operand of the logical operation to operation result.
The patch also simplifies an assume of a constraint of the form: "(exp comparison_op expr) != 0" to true into an assume of "exp comparison_op expr" to true. (And similarly, an assume of the form "(exp comparison_op expr) == 0" to true as an assume of exp comparison_op expr to false.) which improves precision overall.
https://reviews.llvm.org/D22862
Richard Smith [Fri, 23 Dec 2016 02:10:11 +0000 (02:10 +0000)]
Move generation of injected template arguments for a template parameter list
out of an internal function and into ASTContext; this is needed in template
argument deduction for P0522R0.
Richard Smith [Fri, 23 Dec 2016 02:00:24 +0000 (02:00 +0000)]
Only substitute into type of non-type template parameter once, rather than
twice, in finalization of template argument deduction.
This is a re-commit of r290310 (reverted in r290329); the bug found by the
buildbots was fixed in r290399 (we would sometimes build a deduced template
argument with a bogus type).
Richard Smith [Fri, 23 Dec 2016 01:30:39 +0000 (01:30 +0000)]
When merging two deduced non-type template arguments for the same parameter,
fail the merge if the arguments have different types (except if one of them was
deduced from an array bound, in which case take the type from the other).
This is correct because (except in the array bound case) the type of the
template argument in each deduction must match the type of the parameter, so at
least one of the two deduced arguments must have a mismatched type.
This is necessary because we would otherwise lose the type information for the
discarded template argument in the merge, and fail to diagnose the mismatch.
In order to power this, we now properly retain the type of a deduced non-type
template argument deduced from a declaration, rather than giving it the type of
the template parameter; we'll convert it to the template parameter type when
checking the deduced arguments.
Chandler Carruth [Fri, 23 Dec 2016 01:24:49 +0000 (01:24 +0000)]
Cleanup the handling of noinline function attributes, -fno-inline,
-fno-inline-functions, -O0, and optnone.
These were really, really tangled together:
- We used the noinline LLVM attribute for -fno-inline
- But not for -fno-inline-functions (breaking LTO)
- But we did use it for -finline-hint-functions (yay, LTO is happy!)
- But we didn't for -O0 (LTO is sad yet again...)
- We had weird structuring of CodeGenOpts with both an inlining
enumeration and a boolean. They interacted in weird ways and
needlessly.
- A *lot* of set smashing went on with setting these, and then got worse
when we considered optnone and other inlining-effecting attributes.
- A bunch of inline affecting attributes were managed in a completely
different place from -fno-inline.
- Even with -fno-inline we failed to put the LLVM noinline attribute
onto many generated function definitions because they didn't show up
as AST-level functions.
- If you passed -O0 but -finline-functions we would run the normal
inliner pass in LLVM despite it being in the O0 pipeline, which really
doesn't make much sense.
- Lastly, we used things like '-fno-inline' to manipulate the pass
pipeline which forced the pass pipeline to be much more
parameterizable than it really needs to be. Instead we can *just* use
the optimization level to select a pipeline and control the rest via
attributes.
Sadly, this causes a bunch of churn in tests because we don't run the
optimizer in the tests and check the contents of attribute sets. It
would be awesome if attribute sets were a bit more FileCheck friendly,
but oh well.
I think this is a significant improvement and should remove the semantic
need to change what inliner pass we run in order to comply with the
requested inlining semantics by relying completely on attributes. It
also cleans up tho optnone and related handling a bit.
One unfortunate aspect of this is that for generating alwaysinline
routines like those in OpenMP we end up removing noinline and then
adding alwaysinline. I tried a bunch of other approaches, but because we
recompute function attributes from scratch and don't have a declaration
here I couldn't find anything substantially cleaner than this.
Chandler Carruth [Fri, 23 Dec 2016 00:23:01 +0000 (00:23 +0000)]
Make '-disable-llvm-optzns' an alias for '-disable-llvm-passes'.
Much to my surprise, '-disable-llvm-optzns' which I thought was the
magical flag I wanted to get at the raw LLVM IR coming out of Clang
deosn't do that. It still runs some passes over the IR. I don't want
that, I really want the *raw* IR coming out of Clang and I strongly
suspect everyone else using it is in the same camp.
There is actually a flag that does what I want that I didn't know about
called '-disable-llvm-passes'. I suspect many others don't know about it
either. It both does what I want and is much simpler.
This removes the confusing version and makes that spelling of the flag
an alias for '-disable-llvm-passes'. I've also moved everything in Clang
to use the 'passes' spelling as it seems both more accurate (*all* LLVM
passes are disabled, not just optimizations) and much easier to remember
and spell correctly.
This is part of simplifying how Clang drives LLVM to make it cleaner to
wire up to the new pass manager.
Sean Callanan [Thu, 22 Dec 2016 20:03:14 +0000 (20:03 +0000)]
Testbed and skeleton of a new expression parser
Recommitted after formal approval.
LLVM's JIT is now the foundation of dynamic-compilation features for many languages. Clang also has low-level support for dynamic compilation (ASTImporter and ExternalASTSource, notably). How the compiler is set up for dynamic parsing is generally left up to individual clients, for example LLDB's C/C++/Objective-C expression parser and the ROOT project.
Although this arrangement offers external clients the flexibility to implement dynamic features as they see fit, the lack of an in-tree client means that subtle bugs can be introduced that cause regressions in the external clients but aren't caught by tests (or users) until much later. LLDB for example regularly encounters complicated ODR violation scenarios where it is not immediately clear who is at fault.
Other external clients (notably, Cling) rely on similar functionality, and another goal is to break this functionality up into composable parts so that any client can be built easily on top of Clang without requiring extensive additional code.
I propose that the parts required to build a simple expression parser be added to Clang. Initially, I aim to have the following features:
A piece that looks up external declarations from a variety of sources (e.g., from previous dynamic compilations, from modules, or from DWARF) and uses clear conflict resolution rules to reconcile differences, with easily understood errors. This functionality will be supported by in-tree tests.
A piece that works hand in hand with the LLVM JIT to resolve the locations of external declarations so that e.g. variables can be redeclared and (for high-performance applications like DTrace) external variables can be accessed directly from the registers where they reside.
This commit adds a tester that parses a sequence of source files and then uses them as source data for an expression. External references are resolved using an ExternalASTSource that responds to name queries using an ASTImporter. This is the setup that LLDB uses, and the motivating reason for MinimalImport in ASTImporter. When complete, this tester will implement the first of the above goals.
Alexey Bataev [Thu, 22 Dec 2016 19:44:05 +0000 (19:44 +0000)]
[OPENMP] Fix for PR31417: assert failure when compiling trivial openmp
program
Offload related code is not quite ready yet, but some simple examples
must not crash the compiler. Patch fixes the problem in offloading code
with exceptions.
Devin Coughlin [Thu, 22 Dec 2016 17:52:57 +0000 (17:52 +0000)]
[analyzer] Update GTestChecker to tighten API detection
Update the GTestChecker to tighten up the API detection and make it
cleaner in response to post-commit feedback. Also add tests for when
temporary destructors are enabled to make sure we get the expected behavior
when inlining constructors for temporaries.
Artem Dergachev [Thu, 22 Dec 2016 14:48:52 +0000 (14:48 +0000)]
[analyzer] Improve suppress-on-sink behavior in incomplete analyses.
Warnings with suppress-on-sink are discarded during FlushReports when
BugReporter notices that all paths in ExplodedGraph that pass through the
warning eventually run into a sink node.
However, suppress-on-sink fails to filter out false positives when the analysis
terminates too early - by running into analyzer limits, such as block count
limits or graph size limits - and the interruption hits the narrow window
between throwing the leak report and reaching the no-return function call. In
such case the report is there, however suppression-on-sink doesn't work, because
the sink node was never constructed in the incomplete ExplodedGraph.
This patch implements a very partial solution: also suppress reports thrown
against a statement-node that corresponds to a statement that belongs to a
no-return block of the CFG.
[CrashReproducer] Add support for merging -ivfsoverlay
Merge all VFS mapped files inside -ivfsoverlay inputs into the vfs
overlay provided by the crash reproducer. This is the last missing piece
to allow crash reproducers to fully work with user frameworks; when
combined with headermaps, it allows clang to find additional frameworks.
Antonio Maiorano [Thu, 22 Dec 2016 05:10:07 +0000 (05:10 +0000)]
Make FormatStyle.GetStyleOfFile test work on MSVC
Modify getStyle to use vfs::FileSystem::makeAbsolute just like FS.addFile does,
rather than sys::fs::make_absolute. The latter gets the CWD from the platform,
while the former expects it to be set by the client, causing a mismatch when
converting relative paths to absolute.
Sema: print qualified name for overload candidates
Print the fully qualified names for the overload candidates. This makes
it easier to tell what the ambiguity is. Especially if a template
is instantiated after a using namespace, it will not inherit the
namespace where it was declared. The specialization will give a message
about a partial order being ambiguous for the same (unqualified) name,
which does not help identify the failure.
The parameter to ParsePICOpts passed the effective triple and then used
that in a few places and used the actual triple in others. This was
slightly confusing. Rename the parameter to make it more obvious.
This is a recommit of r290149, which was reverted in r290169 due to msan
failures. msan was failing because we were calling
`isMostDerivedAnUnsizedArray` on an invalid designator, which caused us
to read uninitialized memory. To fix this, the logic of the caller of
said function was simplified, and we now have a `!Invalid` assert in
`isMostDerivedAnUnsizedArray`, so we can catch this particular bug more
easily in the future.
Fingers crossed that this patch sticks this time. :)
Original commit message:
This patch does three things:
- Gives us the alloc_size attribute in clang, which lets us infer the
number of bytes handed back to us by malloc/realloc/calloc/any user
functions that act in a similar manner.
- Teaches our constexpr evaluator that evaluating some `const` variables
is OK sometimes. This is why we have a change in
test/SemaCXX/constant-expression-cxx11.cpp and other seemingly
unrelated tests. Richard Smith okay'ed this idea some time ago in
person.
- Uniques some Blocks in CodeGen, which was reviewed separately at
D26410. Lack of uniquing only really shows up as a problem when
combined with our new eagerness in the face of const.
Richard Smith [Wed, 21 Dec 2016 21:42:57 +0000 (21:42 +0000)]
Perform type-checking for a converted constant expression in a template
argument even if the expression is value-dependent (we need to suppress the
final portion of the narrowing check, but the rest of the checking can still be
done eagerly).
This affects template template argument validity and partial ordering under
p0522r0.
Richard Smith [Wed, 21 Dec 2016 01:57:02 +0000 (01:57 +0000)]
Fix defaulted-functions-in-C++98 extension to give the functions the same
effect they would have in C++11. In particular, they do not prevent
value-initialization from performing zero-initialization, nor do they prevent a
struct from being an aggregate.
Richard Smith [Wed, 21 Dec 2016 01:31:56 +0000 (01:31 +0000)]
[c++1z] When initializing a const-qualified class type, don't forget to add on
the requested cv-qualifiers after construction. This usually doesn't matter,
but it does matter within a ?: operator.
Richard Smith [Wed, 21 Dec 2016 01:10:31 +0000 (01:10 +0000)]
Factor out checking of template arguments after deduction into a separate
function. (This change would also allow us to handle default template arguments
in partial specializations if the standard ever permits them.)
Graydon Hoare [Wed, 21 Dec 2016 00:24:39 +0000 (00:24 +0000)]
[modules] Handle modules with nonstandard names in module.private.modulemaps
Summary:
The module system supports accompanying a primary module (say Foo) with
an auxiliary "private" module (defined in an adjacent module.private.modulemap
file) that augments the primary module when associated private headers are
available. The feature is intended to be used to augment the primary
module with a submodule (say Foo.Private), however some users in the wild
are choosing to augment the primary module with an additional top-level module
with a "similar" name (in all cases so far: FooPrivate).
This "works" when a user of the module initially imports a private header,
such as '#import "Foo/something_private.h"' since the Foo import winds up
importing FooPrivate in passing. But if the import is subsequently recorded
in a PCH file, reloading the PCH will fail to validate because of a cross-check
that attempts to find the module.modulemap (or module.private.modulemap) using
HeaderSearch algorithm, applied to the "FooPrivate" name. Since it's stored in
Foo.framework/Modules, not FooPrivate.framework/Modules, the check fails and
the PCH is rejected.
This patch adds a compensatory workaround in the HeaderSearch algorithm
when searching (and failing to find) a module of the form FooPrivate: the
name used to derive filesystem paths is decoupled from the module name
being searched for, and if the initial search fails and the module is
named "FooPrivate", the filesystem search name is altered to remove the
"Private" suffix, and the algorithm is run a second time (still looking for
a module named FooPrivate, but looking in directories derived from Foo).
Accompanying this change is a new warning that triggers when a user loads
a module.private.modulemap that defines a top-level module with a different
name from the top-level module defined in its adjacent module.modulemap.