Richard Smith [Wed, 12 Jun 2013 23:38:09 +0000 (23:38 +0000)]
Simplify: we don't need any special-case lifetime extension when initializing
declarations of reference type; they're handled by the general case handling of
MaterializeTemporaryExpr.
Richard Smith [Wed, 12 Jun 2013 22:31:48 +0000 (22:31 +0000)]
PR12086, PR15117
Introduce CXXStdInitializerListExpr node, representing the implicit
construction of a std::initializer_list<T> object from its underlying array.
The AST representation of such an expression goes from an InitListExpr with a
flag set, to a CXXStdInitializerListExpr containing a MaterializeTemporaryExpr
containing an InitListExpr (possibly wrapped in a CXXBindTemporaryExpr).
This more detailed representation has several advantages, the most important of
which is that the new MaterializeTemporaryExpr allows us to directly model
lifetime extension of the underlying temporary array. Using that, this patch
*drastically* simplifies the IR generation of this construct, provides IR
generation support for nested global initializer_list objects, fixes several
bugs where the destructors for the underlying array would accidentally not get
invoked, and provides constant expression evaluation support for
std::initializer_list objects.
Richard Smith [Wed, 12 Jun 2013 21:51:50 +0000 (21:51 +0000)]
Move detection of reference members binding to temporaries from building of
CXXCtorInitializers to the point where we perform the questionable lifetime
extension. This exposed a selection of false negatives in the warning.
Richard Trieu [Wed, 12 Jun 2013 21:20:57 +0000 (21:20 +0000)]
Introducing -Wheader-guard, a warning that checks header guards actually work
properly. This warning checks that the #ifndef and #define directives at
the beginning of a header refer to the same macro name. Includes a fix-it
hint to correct the header guard.
Richard Smith [Wed, 12 Jun 2013 20:42:33 +0000 (20:42 +0000)]
Reapply r183721, reverted in r183776, with a fix for a bug in the former (we
were lacking ExprWithCleanups nodes in some cases where the new approach to
lifetime extension needed them).
Original commit message:
Rework IR emission for lifetime-extended temporaries. Instead of trying to walk
into the expression and dig out a single lifetime-extended entity and manually
pull its cleanup outside the expression, instead keep a list of the cleanups
which we'll need to emit when we get to the end of the full-expression. Also
emit those cleanups early, as EH-only cleanups, to cover the case that the
full-expression does not terminate normally. This allows IR generation to
properly model temporary lifetime when multiple temporaries are extended by the
same declaration.
We have a pre-existing bug where an exception thrown from a temporary's
destructor does not clean up lifetime-extended temporaries created in the same
expression and extended to automatic storage duration; that is not fixed by
this patch.
Pavel Labath [Wed, 12 Jun 2013 07:45:04 +0000 (07:45 +0000)]
Fix memory corruption in CStringChecker
Summary:
"register" functions for the checker were caching the checker objects in a
static variable. This caused problems when the function is called with a
different CheckerManager.
Eli Friedman [Tue, 11 Jun 2013 21:48:11 +0000 (21:48 +0000)]
Correctly handle designated initializers which modify an array initialized
with a string. This case is sort of tricky because we can't modify the
StringLiteral used to represent such initializers.
We are forced to decompose the string into individual characters.
Samuel Benzaquen [Tue, 11 Jun 2013 18:51:07 +0000 (18:51 +0000)]
Reduce the number of symbols on the object file.
Summary:
Some compilers where failing with this file because the number of symbols was above 2**15.
- Replace std::list<> and std::vector<> with plain arrays.
- Change VariadicMatcherCreateCallback to be a function template, and a
single class that wraps the instantiations.
- Remove some more unnecessary template arguments and function calls.
Insert a space at the start of a line comment in case it starts with an alphanumeric character.
Summary:
"//Test" becomes "// Test". This change is aimed to improve code
readability and conformance to certain coding styles. If a comment starts with a
non-alphanumeric character, the space isn't added, e.g. "//-*-c++-*-" stays
unchanged.
Richard Smith [Tue, 11 Jun 2013 02:41:00 +0000 (02:41 +0000)]
Rework IR emission for lifetime-extended temporaries. Instead of trying to walk
into the expression and dig out a single lifetime-extended entity and manually
pull its cleanup outside the expression, instead keep a list of the cleanups
which we'll need to emit when we get to the end of the full-expression. Also
emit those cleanups early, as EH-only cleanups, to cover the case that the
full-expression does not terminate normally. This allows IR generation to
properly model temporary lifetime when multiple temporaries are extended by the
same declaration.
We have a pre-existing bug where an exception thrown from a temporary's
destructor does not clean up lifetime-extended temporaries created in the same
expression and extended to automatic storage duration; that is not fixed by
this patch.
Eli Friedman [Tue, 11 Jun 2013 01:08:22 +0000 (01:08 +0000)]
Fix a FIXME in a testcase about packed structs and calls I left around
while fixing a related bug. The fix here was simpler than I thought it
would be.
[libclang] Allow building a precompiled preamble with compiler errors
A while ago we allowed libclang to build a PCH that had compiler errors; this was to retain the performance
afforded by a PCH even if the user's code is in an intermediate state.
Adrian Prantl [Mon, 10 Jun 2013 21:36:55 +0000 (21:36 +0000)]
cleanup (address some more review comments for r183474):
- reduce default buffer size to 64, which will still be large enough to
hold any property names found in the wild.
- get rid of the /*static*/ comments.
Jordan Rose [Mon, 10 Jun 2013 19:34:30 +0000 (19:34 +0000)]
[analyzer] SATestBuild: Don't require reference results to have logs.
The Logs directory isn't used for testing, so it's filtered out ahead of
time. However, there's then no reason to include it in version control at
all. Don't error if it's not present.
Richard Trieu [Mon, 10 Jun 2013 18:52:07 +0000 (18:52 +0000)]
Add a new warning, -Wlogical-not-parentheses, to -Wparentheses.
This warning triggers on the logical not of a non-boolean expression on the
left hand side of comparison. Often, the user meant to negate the comparison,
not just the left hand side of the comparison. Two notes are also emitted,
the first with a fix-it to add parentheses around the comparison, and the other
to put parenthesis around the not expression to silence the warning.
bool not_equal(int x, int y) {
return !x == y; // warn here
}
return !(x == y); // first fix-it, to negate comparison.
return (!x) == y; // second fix-it, to silence warning.
Reid Kleckner [Sun, 9 Jun 2013 16:45:02 +0000 (16:45 +0000)]
[CodeGen] Move EHScopeStack to CGCleanup.h from CodeGenFunction.h
No functionality change. CGCleanup.cpp provides the implementation for
EHScopeStack, so it seems more consistent to place the class definition
in CGCleanup.h.
This should also help solve a header ordering problem that I have.
Faisal Vali [Sat, 8 Jun 2013 19:47:52 +0000 (19:47 +0000)]
Fix the parser's updating of the template depth when parsing local templates and late-parsed templates.
This is a slight tweak of r180708; It avoids incrementing depth when non-template local classes nested within member templates of local classes are encountered.
This patch was LGTM'd by Doug http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20130506/079656.html and passed the regression tests that normally pass (i.e. excluding many Module and Index tests on Windows that fail regardless)
Faisal Vali [Sat, 8 Jun 2013 19:39:00 +0000 (19:39 +0000)]
Revert r183618.
I ran clang-format on my patch but it seemed to have wreaked havoc with new lines - might have to do with using it on windows :( will resubmit once i've cleaned this issue up. sorry.
Faisal Vali [Sat, 8 Jun 2013 19:33:09 +0000 (19:33 +0000)]
Fix the parser's updating of the template depth when parsing local templates and late-parsed templates. This is a slight tweak of r180708; It avoids incrementing depth when non-template local classes nested within member templates of local classes are encountered.
This patch was LGTM'd by Doug http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20130506/079656.html and passed the regression tests that normally pass (i.e. excluding many Module and Index tests on Windows that fail regardless)
Reid Kleckner [Sat, 8 Jun 2013 17:28:56 +0000 (17:28 +0000)]
[Sema] Make FunctionType's TSI use unadjusted argument types
This helps preserve the type-as-written in the AST, which we need for
MSVC mangling. In particular, we need to preserve the types of array
parameters in function pointer types.
The essence of this change is:
- QualType ArgTy = Param->getType();
+ QualType ArgTy = Param->getTypeSourceInfo()->getType();
... followed by the adjustment in ActOnFunctionDeclarator().
Anna Zaks [Sat, 8 Jun 2013 00:29:24 +0000 (00:29 +0000)]
[analyzer; alternate edges] Fix the edge locations in presence of macros.
We drew the diagnostic edges to wrong statements in cases the note was on a macro.
The fix is simple, but seems to work just fine for a whole bunch of test cases (plist-macros.cpp).
Also, removes an unnecessary edge in edges-new.mm, when function signature starts with a macro.
Adrian Prantl [Sat, 8 Jun 2013 00:16:55 +0000 (00:16 +0000)]
Debug info: An if condition now creates a lexical scope of its own.
Two variables with the same name declared in two if conditions in the same
scope are no longer coalesced into one.
Adrian Prantl [Fri, 7 Jun 2013 22:29:12 +0000 (22:29 +0000)]
address some comments on r183474:
- factor the name construction part out from constructSetterName
- rename constructSetterName to the more appropriate constructSetterSelector
no functionality change intended.
rdar://problem/14035789
Fixed calculation of penalty when breaking tokens.
Summary:
Introduced two new style parameters: PenaltyBreakComment and
PenaltyBreakString. Add penalty for each character of a breakable token beyond
the column limit (this relates mainly to comments, as they are broken only on
whitespace). Tuned PenaltyBreakComment to prefer comment breaking over breaking
inside most binary expressions.
Fixed a bug that prevented *, & and && from being considered TT_BinaryOperator
in the presense of adjacent comments.
Richard Smith [Fri, 7 Jun 2013 02:33:37 +0000 (02:33 +0000)]
PR16243: Use CXXThisOverride during template instantiation, and fix up the
places which weren't setting it up properly. This allows us to get the right
cv-qualifiers for 'this' when it appears outside a method body in a class
template.
blocks: fixes an ast bug when block pointer variable
is evaluated in a condition expression and then
dereferenced to envoke the block. This is
pr15663 and I applied a slight variation of the
patch with a test case. (patch is from
Arthur O'Dwyer). Also // rdar://14085217
Tim Northover [Fri, 7 Jun 2013 00:04:50 +0000 (00:04 +0000)]
Diagnose malformed x86 inline asm using 'y' constraint.
X86's 'y' inline assembly constraint represents an MMX register, this change
prevents Clang from hitting an assertion when passed an incompatible type to
deal with.
Eli Friedman [Fri, 7 Jun 2013 00:04:31 +0000 (00:04 +0000)]
Perform dynamic alignment computations so that the data in TypeLocs is
correctly aligned. Not performing such computations led to misaligned loads,
which crash on some platforms and are generally bad on other platforms.
The implementation of TypeLocBuilder::pushImpl is rather messy; code using
TypeLocBuilder accidentally assumes that partial TypeLocs are
laid out like a complete TypeLoc. As a followup, I intend to work on
fixing the TypeLocBuilder API to avoid exposing partial TypeLocs; this should
substantially simplify the implemementation.
projects / clang / log