Bob Wilson [Fri, 23 Nov 2012 06:14:39 +0000 (06:14 +0000)]
Reapply a subset of r167567 to clean up Darwin-specific code for invoking gcc.
Unlike my previous attempt at this, this patch leaves intact the check for
whether clang can handle the input file type, and for non-Darwin toolchains it
will invoke gcc for things it cannot handle. For Darwin toolchains, the
behavior reported in pr14338 still occurs with this patch, but that is a
definite improvement from what happens currently, where it just crashes with
an assertion failure.
Richard Smith [Thu, 22 Nov 2012 00:24:47 +0000 (00:24 +0000)]
Fix CXXRecordDecl::forallBases to not look through bases which are dependent
and defined within the current instantiation, but which are not part of the
current instantiation. Previously, it would look at bases which could be
specialized separately from the current template.
Chandler Carruth [Wed, 21 Nov 2012 23:40:23 +0000 (23:40 +0000)]
Fix the '-fuse-init-array' option to actually be an option.
Previously, this flag to CC1 was never exposed at the clang driver
layer, and if you happened to enable it (by being on Android or GCC 4.7
platform), you couldn't *disable* it, because there was no 'no' variant.
The whole thing was confusingly implemented.
Now, the target-specific flag processing gets the driver arg list, and
we use standard hasFlag with a default based on the GCC version and/or
Android platform. The user can still pass the 'no-' variant to forcibly
disable the flag, or pass the positive variant to clang itself to enable
the flag.
The test has also been substantially cleaned up and extended to cover
these use cases.
Alexey Samsonov [Wed, 21 Nov 2012 14:17:42 +0000 (14:17 +0000)]
[Sanitizer] force linking with static sanitizer runtimes on Darwin even if they are not found in resource directory. Add test checking sanitizer linker flags on Darwin.
Rafael Espindola [Tue, 20 Nov 2012 23:23:57 +0000 (23:23 +0000)]
Don't walk a linked list twice in the same function. On my machine this takes
"clang -cc1" on a file with 10k repetitions of
extern int no_such_variable;
from 1.434s to 1.133s.
Kaelyn Uhrain [Mon, 19 Nov 2012 18:49:53 +0000 (18:49 +0000)]
When adding a NamedDecl to a correction, add the underlying Decl (via
getUnderlyingDecl()) so that derivatives of
CorrectionCandidateCallback::ValidateCandidate(...) don't have to worry
about being thrown by UsingDecls and such.
Logan Chien [Mon, 19 Nov 2012 11:45:31 +0000 (11:45 +0000)]
Enable -fuse-init-array for Android X86/MIPS.
The dynamic linker of Android does not support .ctors/.dtors.
We should emit .init_array and .fini_array regardless the
gcc version.
NOTE: This patch does not affect the ARM backend, because
it is required to generate .init_array and .fini_array
for program targeting ARM AAPCS and AEABI.
Chandler Carruth [Mon, 19 Nov 2012 03:52:03 +0000 (03:52 +0000)]
Completely re-work how the Clang driver interprets PIC and PIE options.
There were numerous issues here that were all entangled, and so I've
tried to do a general simplification of the logic.
1) The logic was mimicing actual GCC bugs, rather than "features". These
have been fixed in trunk GCC, and this fixes Clang as well. Notably,
the logic was always intended to be last-match-wins like any other
flag.
2) The logic for handling '-mdynamic-no-pic' was preposterously unclear.
It also allowed the use of this flag on non-Darwin platforms where it
has no actual meaning. Now this option is handled directly based on
tests of how llvm-gcc behaves, and it is only supported on Darwin.
3) The APIs for the Driver's ToolChains had the implementation ugliness
of dynamic-no-pic leaking through them. They also had the
implementation details of the LLVM relocation model flag names
leaking through.
4) The actual results of passing these flags was incorrect on Darwin in
many cases. For example, Darwin *always* uses PIC level 2 if it uses
in PIC level, and Darwin *always* uses PIC on 64-bit regardless of
the flags specified, including -fPIE. Darwin never compiles in PIE
mode, but it can *link* in PIE mode.
5) Also, PIC was not always being enabled even when PIE was. This isn't
a supported mode at all and may have caused some fallout in builds
with complex PIC and PIE interactions.
The result is (I hope) cleaner and clearer for readers. I've also left
comments and tests about some of the truly strage behavior that is
observed on Darwin platforms. We have no real testing of Windows
platforms and PIC, but I don't have the tools handy to figure that out.
Hopefully others can beef up our testing here.
Unfortunately, I can't test this for every platform. =/ If folks have
dependencies on these flags that aren't covered by tests, they may
break. I've audited and ensured that all the changes in behavior of the
existing tests are intentional and good. In particular I've tried to
make sure the Darwin behavior (which is more suprising than the Linux
behavior) also matches that of 'gcc' on my mac.
objective-C: Do not issue deprecated warning about implementation
of a deprecated method in original class (or category), only
in overrides. // rdar://12717705
Andy Gibbs [Sat, 17 Nov 2012 19:18:27 +0000 (19:18 +0000)]
Prevent premature macro expansion in __has_builtin, __has_feature,
__has_attribute, __has_extension, making them behave more akin to
conventional macros.
Andy Gibbs [Sat, 17 Nov 2012 19:15:38 +0000 (19:15 +0000)]
Refactored duplicate string literal lexing code within Preprocessor, into a
common LexStringLiteral function. In doing so, some consistency problems have
been ironed out (e.g. where the first token in the string literal was lexed
with macro expansion, but subsequent ones were not) and also an erroneous
diagnostic has been corrected.
LexStringLiteral is complemented by a FinishLexStringLiteral function which
can be used in the situation where the first token of the string literal has
already been lexed.
Ted Kremenek [Sat, 17 Nov 2012 02:00:00 +0000 (02:00 +0000)]
Switch -Wuninitialized to use a reverse-post order traversal as
an initial baseline for enqueued blocks, but use a simple DFS stack
for propagating changes quickly up back edges.
This provides a 3.5% reduction in -fsyntax-only time on sqlite3.c.
Chad Rosier [Fri, 16 Nov 2012 22:31:39 +0000 (22:31 +0000)]
[driver] Add the missing TY_PP_ObjCXX_Alias case to the isCXX function.
This was causing different behavior when using -x objective-c++-cpp-output as
compared to -x objc++-cpp-output. Specifically, the driver was not adding the
-fcxx-exceptions flag in the latter case.
rdar://12680312
Dmitri Gribenko [Fri, 16 Nov 2012 21:43:31 +0000 (21:43 +0000)]
StmtDumper: remove incomplete support for limiting the maximum dump depth.
There are better ways of limiting the amount of information if there is a need
for that.
Daniel Jasper [Fri, 16 Nov 2012 18:39:22 +0000 (18:39 +0000)]
Fix partial-match-bind-behavior with forEachDescendant() matchers.
The problem is that a partial match of an (explicit or implicit) allOf matcher
binds results, i.e.
recordDecl(decl().bind("x"), hasName("A"))
can very well bind a record that is not named "A". With this fix, the common
cases of stumbling over this bug are fixed by the BoundNodesMap overwriting the
results of a partial match. An error can still be created with a weird
combination of anyOf and allOf (see inactive test). We need to decide whether
this is worth fixing, as the fix will have performance impact.
This has been broken for a while, but the branch was never being taken.
(We were trying to do 'str + floatVal'; now we do 'str % intVal' and use
the '%d' format.)
[libclang] When caching code-completion results, pass the CachedCompletionAllocator
to the CodeCompletionTUInfo that is going to be used to get the results.
Previously we would use ASTUnit's CodeCompletionTUInfo which has its own allocator
that will go away when we reparse. That could result in a use-after-free bug when
getting the parent context name from a CodeCompletionString.
Daniel Dunbar [Fri, 16 Nov 2012 01:51:11 +0000 (01:51 +0000)]
CPP Output: Do not emit an enter file marker for the main file.
- This diverges from gcc, and confuses tools (like dtrace) which track # line
markers as a way to determine which content is in the context of the main
file.
Richard Trieu [Fri, 16 Nov 2012 01:32:40 +0000 (01:32 +0000)]
Take into account the zero sign bit for positive numbers when computing the bit
width of an enum with negative values in IntRange. Include a test for
-Wtautological-constant-out-of-range-compare where this had manifested.
Richard Smith [Fri, 16 Nov 2012 00:53:38 +0000 (00:53 +0000)]
A step towards sorting out handling of triviality of special members in C++11.
Separate out the notions of 'has a trivial special member' and 'has a
non-trivial special member', and use them appropriately. These are not
opposites of one another (there might be no special member, or in C++11 there
might be a trivial one and a non-trivial one). The CXXRecordDecl predicates
continue to produce incorrect results, but do so in fewer cases now, and
they document the cases where they might be wrong.
No functionality changes are intended here (they will come when the predicates
start producing the right answers...).
Richard Smith [Thu, 15 Nov 2012 22:54:20 +0000 (22:54 +0000)]
PR9903: Recover from a member functon declared with the 'typedef' specifier by
dropping the specifier, just like we do for non-member functions and function
templates declared 'typedef'. Patch by Brian Brooks!
Jordan Rose [Thu, 15 Nov 2012 20:10:05 +0000 (20:10 +0000)]
[analyzer] Fix a use-after-free introduced in r168019.
In code like this:
void foo() {
bar();
baz();
}
...the location for the call to 'bar()' was being used as a backup location
for the call to 'baz()'. This is fine unless the call to 'bar()' is deemed
uninteresting and that part of the path deleted.
(This looks like a logic error as well, but in practice the only way 'baz()'
could have an invalid location is if the entire body of 'foo()' is
synthesized, meaning the call to 'bar()' will be using the location of the
call to 'foo()' anyway. Nevertheless, the new version better matches the
intent of the code.)
Found by Matt Beaumont-Gay using ASan. Thanks, Matt!
Jordan Rose [Thu, 15 Nov 2012 19:11:43 +0000 (19:11 +0000)]
[analyzer] Report leaks at the closing brace of a function body.
This fixes a few cases where we'd emit path notes like this:
+---+
1| v
p = malloc(len);
^ |2
+---+
In general this should make path notes more consistent and more correct,
especially in cases where the leak happens on the false branch of an if
that jumps directly to the end of the function. There are a couple places
where the leak is reported farther away from the cause; these are usually
cases where there are several levels of nested braces before the end of
the function. This still matches our current behavior for when there /is/
a statement after all the braces, though.
Jordan Rose [Thu, 15 Nov 2012 19:11:27 +0000 (19:11 +0000)]
[analyzer] Mark symbol values as dead in the environment.
This allows us to properly remove dead bindings at the end of the top-level
stack frame, using the ReturnStmt, if there is one, to keep the return value
live. This in turn removes the need for a check::EndPath callback in leak
checkers.
This does cause some changes in the path notes for leak checkers. Previously,
a leak would be reported at the location of the closing brace in a function.
Now, it gets reported at the last statement. This matches the way leaks are
currently reported for inlined functions, but is less than ideal for both.
block extended signatur option. Change previous option
to a cc1 -fencode-extended-block-signature and pass it
to cc1 and recognize this option to produce extended block
type signature. // rdar://12109031
Benjamin Kramer [Thu, 15 Nov 2012 15:18:42 +0000 (15:18 +0000)]
Do not cache a pointer to ExprEvalContexts.back().
It may become a dangling pointer if the underlying SmallVector reallocates.
Sadly the testcase is really large and doesn't reduce well because of
SmallVector's reallocation patterns.
Lang Hames [Thu, 15 Nov 2012 07:51:26 +0000 (07:51 +0000)]
Make -ffp-contract a codegen option, rather than a laguage option. This makes
more sense anyway - it determines how expressions are codegen'd. It also ensures
that -ffp-contract=fast has the intended effect when compiling LLVM IR.
Jordan Rose [Thu, 15 Nov 2012 02:07:23 +0000 (02:07 +0000)]
[analyzer] Make sure calls in synthesized functions have valid path locations.
We do this by using the "most recent" good location: if a synthesized
function 'A' calls another function 'B', the path notes for the call to 'B'
will be placed at the same location as the path note for calling 'A'.
Similarly, the call to 'A' will have a note saying "Entered call from...",
and now we just don't emit that (since the user doesn't have a body to look
at anyway).
Previously, we were doing this for the "Calling..." notes, but not for the
"Entered call from..." or "Returning to caller". This caused a crash when
the path entered and then exiting a call within a synthesized body.