Douglas Gregor [Thu, 3 Feb 2011 23:41:12 +0000 (23:41 +0000)]
If any Fix-Its attached to a diagnostic have invalid source locations
or source locations that refer into a macro instantiation, delete all
of the Fix-Its on that diagnostic.
Oscar Fuentes [Thu, 3 Feb 2011 20:57:53 +0000 (20:57 +0000)]
Optionally, Clang now builds using LLVM as an external library.
The user sets the variable CLANG_PATH_TO_LLVM_BUILD pointing to a LLVM
build directory or to a directory where LLVM was installed. When using
a non-installed LLVM build, it is necessary to set
CLANG_PATH_TO_LLVM_SOURCE as well.
The difference with gcc is that it warns if you overload virtual methods only if
the method doesn't also override any method. This is to cut down on the number of warnings
and make it more useful like reported here: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20423.
If we want to warn that not all overloads are overriden we can have an additional
warning like -Wpartial-override.
-Woverloaded-virtual, unlike gcc, is added to -Wmost. Addresses rdar://8757630.
Douglas Gregor [Thu, 3 Feb 2011 17:17:35 +0000 (17:17 +0000)]
Teach SourceManager::getLocation() how to cope with a source file
whose inode has changed since the file was first created and that is
being seen through a different path name (e.g., due to symlinks or
relative path elements), such that its FileEntry pointer doesn't match
a known FileEntry pointer. Since this requires a system call (to
stat()), we only perform this deeper checking if we can't find the
file by comparing FileEntry pointers.
Also, add a micro-optimization where we don't bother to compute line
numbers when given the location (1, 1). This improves the
efficiency of clang_getLocationForOffset().
John McCall [Thu, 3 Feb 2011 08:15:49 +0000 (08:15 +0000)]
More capturing of 'this': implicit member expressions. Getting that
right for anonymous struct/union members led to me discovering some
seemingly broken code in that area of Sema, which I fixed, partly by
changing the representation of member pointer constants so that
IndirectFieldDecls aren't expanded. This led to assorted cleanups with
member pointers in CodeGen, and while I was doing that I saw some random
other things to clean up.
Daniel Dunbar [Thu, 3 Feb 2011 03:45:00 +0000 (03:45 +0000)]
Frontend: Switch -header-include-file output to use unbuffered raw_ostreams with
the atomic writes option, since the intent is that this option be set for an
entire build, which may have any number of compiler instances writing to the
same output file.
Anders Carlsson [Thu, 3 Feb 2011 02:08:44 +0000 (02:08 +0000)]
Don't try to mark virtual members referenced for classes where the key function
is not defined in the current translation unit. Doing so lead to compile errors
such as PR9114.
Instead, when CodeGen is building the vtable, don't try to emit a definition
for functions that aren't marked used in the current translation unit.
Fixes PR9114.
Douglas Gregor [Wed, 2 Feb 2011 22:30:17 +0000 (22:30 +0000)]
Canonicalize path names in the file manager before performing a lookup
on that name. Canonicalization eliminates silliness such as "." and
"foo/.." that breaks the uniquing of files in the presence of virtual
files or files whose inode numbers have changed during
parsing/re-parsing. c-index-test isn't able to create this crazy
situation, so I've resorted to testing outside of the Clang
tree. Fixes <rdar://problem/8928220>.
Note that this hackery will go away once we have a real virtual file
system on which we can layer FileManager; the virtual-files hack is
showing cracks.
Daniel Dunbar [Wed, 2 Feb 2011 21:11:35 +0000 (21:11 +0000)]
clang: Add support for a CC_PRINT_HEADERS environment variable, which mirrors
CC_PRINT_OPTIONS and can be used to get some out-of-band information on header
usage from a build.
Douglas Gregor [Wed, 2 Feb 2011 19:04:30 +0000 (19:04 +0000)]
Revert r124704, which uniqued code-completion strings. The space
savings of 25% sounds impressive, except that this amounted to only
about 360k in our standard "large" completion result set (40,000
results). Since code completion is performance-sensitive, the 4%
slowdown due to uniquing outweighs the 360k benefit.
Daniel Dunbar [Wed, 2 Feb 2011 15:41:17 +0000 (15:41 +0000)]
Frontend: Factor out header include dumping (-H) into its own preprocessor
callbacks class.
- Aside from being generally cleaner, this also allows -H to work correctly in
modes other than standard preprocessing (e.g., -c, -MM, etc.)
John McCall [Wed, 2 Feb 2011 13:00:07 +0000 (13:00 +0000)]
An insomniac stab at making block declarations list the variables they close
on, as well as more reliably limiting invalid references to locals from
nested scopes.
Ted Kremenek [Tue, 1 Feb 2011 22:36:09 +0000 (22:36 +0000)]
Don't warn about extraneous '()' around a comparison if it occurs within a macro.
Macros frequently contain extra '()' to make instantiation less error prone.
This warning was flagging a ton of times on postgresql because of its use of macros.
John McCall [Tue, 1 Feb 2011 21:35:06 +0000 (21:35 +0000)]
Do the right thing for zero-initializing VLAs that don't have a zero
bit-pattern. It's not clear that this is actually useful given current
language restrictions.
Douglas Gregor [Tue, 1 Feb 2011 19:23:04 +0000 (19:23 +0000)]
Allocate CodeCompletionString and all of its components in a
BumpPtrAllocator, rather than manually new/delete'ing them. This
optimization also allows us to avoid allocating memory for and copying
constant strings (e.g., "return", "class").
This also required embedding the priority and availability of results
within the code completion string, to avoid extra memory allocation
within libclang.
Warn for "if ((a == b))" where the equality expression is needlessly wrapped inside parentheses.
It's highly likely that the user intended an assignment used as condition.
Ted Kremenek [Tue, 1 Feb 2011 17:43:21 +0000 (17:43 +0000)]
Add temporary hack to -Wuninitialize to create a separate CFG (for C++ code) that doesn't include implicit dtors.
Implicit dtors confuse the ad hoc path-sensitivity of UninitializedValuesV2.cpp. This isn't
the ideal solution, as it will directly impact compile time, but should significantly reduce
the noise of -Wuninitialized on some code bases.
This immediately "fixes" the false positive reported in PR 9063, although this
isn't the right fix in the long run.
Douglas Gregor [Tue, 1 Feb 2011 15:24:58 +0000 (15:24 +0000)]
Fix a thinko where I didn't update a consistency check for
PackExpansionType in the AST reader. We need more testing for variadic
templates + PCH, but this fixes PR9073.
John McCall [Tue, 1 Feb 2011 08:20:08 +0000 (08:20 +0000)]
The code trying to assign a typedef to an anonymous tag declaration was
extremely rambunctious, both on parsing and on template instantiation.
Calm it down, fixing an internal consistency assert on anonymous enum
instantiation manglings.
Set visibility for available_externally globals. This is important for two reasons:
* llvm-link would complains about mismatched visibility
* If we produce a relocation with an available_externally, it is good to know that
it is hidden.
Douglas Gregor [Mon, 31 Jan 2011 18:51:41 +0000 (18:51 +0000)]
Implement reasonable conversion ranking for Objective-C pointer
conversions (<rdar://problem/8592139>) for overload resolution. The
conversion ranking mirrors C++'s conversion ranking fairly closely,
except that we use a same pseudo-subtyping relationship employed by
Objective-C pointer assignment rather than simple checking
derived-to-base conversions. This change covers:
- Conversions to pointers to a specific object type are better than
conversions to 'id', 'Class', qualified 'id', or qualified 'Class'
(note: GCC doesn't perform this ranking, but it matches C++'s rules
for ranking conversions to void*).
- Conversions to qualified 'id' or qualified 'Class' are better than
conversions to 'id' or 'Class', respectively.
- When two conversion sequences convert to the same type, rank the
conversions based on the relationship between the types we're
converting from.
- When two conversion sequences convert from the same non-id,
non-Class type, rank the conversions based on the relationship of
the types we're converting to. (note: GCC allows this ranking even
when converting from 'id', which is extremeley dangerous).
Douglas Gregor [Mon, 31 Jan 2011 16:09:46 +0000 (16:09 +0000)]
Implement the suggested resolution to core issue 547, extended to also
allow ref-qualifiers on function types used as template type
arguments. GNU actually allows cv-qualifiers on function types in many
places where it shouldn't, so we currently categorize this as a GNU
extension.
Error for use of field from anonymous struct or union should say "invalid use of nonstatic data member"
not "call to non-static member function without an object argument".
Amazing that there are still issues with the fields of anonymous struct/unions..
Allow taking the address of such a field for a pointer-to-member constant. Fixes rdar://8818236.