]> granicus.if.org Git - clang/log
clang
12 years agoAdd support for printing integer literals of type short, unsigned short,
Richard Trieu [Mon, 7 Nov 2011 18:40:31 +0000 (18:40 +0000)]
Add support for printing integer literals of type short, unsigned short,
__int128_t and __uint128_t.  Short and unsigned short integer literals support
is only to work around a crasher as reported in PR11179 and will be removed
once Clang no longer builds short integer literals.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143977 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[arcmt] In GC, change '__weak' -> '__unsafe_unretained' when applied
Argyrios Kyrtzidis [Mon, 7 Nov 2011 18:40:29 +0000 (18:40 +0000)]
[arcmt] In GC, change '__weak' -> '__unsafe_unretained' when applied
to objects of classes that don't support ARC weak

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143976 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoWhen applying ARC __weak to a non-objc pointer, do not give error that
Argyrios Kyrtzidis [Mon, 7 Nov 2011 18:40:21 +0000 (18:40 +0000)]
When applying ARC __weak to a non-objc pointer, do not give error that
__weak is unsupported by the deployment target, since it is going to be
ignored anyway.

Makes it easier for incremental migration from GC.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143975 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBetter fix for -Wnon-virtual-dtor
Matt Beaumont-Gay [Mon, 7 Nov 2011 18:30:47 +0000 (18:30 +0000)]
Better fix for -Wnon-virtual-dtor

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143972 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove duplicate listings for core.DivideZero
Matt Beaumont-Gay [Mon, 7 Nov 2011 18:00:03 +0000 (18:00 +0000)]
Remove duplicate listings for core.DivideZero

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143969 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDrastically simplify the mapping from the declaration corresponding to
Douglas Gregor [Mon, 7 Nov 2011 17:43:18 +0000 (17:43 +0000)]
Drastically simplify the mapping from the declaration corresponding to
the injected-class-name of a class (or class template) to the
declaration that results from substituting the given template
arguments. Previously, we would actually perform a substitution into
the injected-class-name type and then retrieve the resulting
declaration. However, in certain, rare circumstances involving
deeply-nested member templates, we would get the wrong substitution
arguments.

This new approach just matches up the declaration with a declaration
that's part of the current context (or one of its parents), which will
either be an instantiation (during template instantiation) or the
declaration itself (during the definition of the template). This is
both more efficient (we're avoiding a substitution) and more correct
(we can't get the template arguments wrong in the member-template
case).

Fixes <rdar://problem/9676205>.

Reinstated, now that we have the fix in r143967.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143968 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTighten up the conditions under which we consider ourselves to be
Douglas Gregor [Mon, 7 Nov 2011 17:33:42 +0000 (17:33 +0000)]
Tighten up the conditions under which we consider ourselves to be
entering the context of a nested-name-specifier. Fixes
<rdar://problem/10397846>.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143967 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAppease -Wnon-virtual-dtor
Matt Beaumont-Gay [Mon, 7 Nov 2011 17:16:31 +0000 (17:16 +0000)]
Appease -Wnon-virtual-dtor

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143965 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix an unsafe use of Twine.
Benjamin Kramer [Mon, 7 Nov 2011 16:02:25 +0000 (16:02 +0000)]
Fix an unsafe use of Twine.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143962 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoConstant expression evaluation: support for arrays.
Richard Smith [Mon, 7 Nov 2011 09:22:26 +0000 (09:22 +0000)]
Constant expression evaluation: support for arrays.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143922 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRip out one of the features I added for the driver-include-management.
Chandler Carruth [Mon, 7 Nov 2011 09:17:31 +0000 (09:17 +0000)]
Rip out one of the features I added for the driver-include-management.
We don't actually need a separate flag for non-sysrooted paths as the
driver has to manage the sysroot anyways. The driver is not infrequently
adding paths to the header search based on their existence on the
filesystem. For that, it has to add the sysroot anyways, we should pass
it on down to CC1 already joined. More importantly, the driver cannot in
all cases distinguish between sysrooted paths and paths that are
relative to the Clang binary's installation directory. Essentially, we
always need to ignore the system root for these internal header search
options. It turns out in most of the places we were already providing
the system root in the driver, and then another one in CC1 so this fixes
several bugs.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143917 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix Linux libc++ usage. Somehow this slipped through during the port.
Chandler Carruth [Mon, 7 Nov 2011 09:01:17 +0000 (09:01 +0000)]
Fix Linux libc++ usage. Somehow this slipped through during the port.

Test cases for this and all the rest of the port are still in the works,
but will wait for a fixed computer and post 3.0 merging...

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143916 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix 32-bit build bots and remove some casting-away-const warnings.
Richard Smith [Mon, 7 Nov 2011 07:31:09 +0000 (07:31 +0000)]
Fix 32-bit build bots and remove some casting-away-const warnings.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143914 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoColorize. (this is consistent with the coloring in diagnostics.html, but perhaps...
David Blaikie [Mon, 7 Nov 2011 06:28:33 +0000 (06:28 +0000)]
Colorize. (this is consistent with the coloring in diagnostics.html, but perhaps that's a bit out of date because it doesn't look like current clang behavior)

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143913 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[analyzer] Add the Checker Developer Manual to the menu.
Anna Zaks [Mon, 7 Nov 2011 05:40:51 +0000 (05:40 +0000)]
[analyzer] Add the Checker Developer Manual to the menu.
It's still under construction but has enough info to be useful.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143912 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[analyzer] Add overview and checker registration to the checker developer manual.
Anna Zaks [Mon, 7 Nov 2011 05:36:29 +0000 (05:36 +0000)]
[analyzer] Add overview and checker registration to the checker developer manual.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143911 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRip out CK_GetObjCProperty.
John McCall [Mon, 7 Nov 2011 05:09:54 +0000 (05:09 +0000)]
Rip out CK_GetObjCProperty.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143910 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoConstant expression evaluation: preserve subobject designator when flattening a
Richard Smith [Mon, 7 Nov 2011 05:07:52 +0000 (05:07 +0000)]
Constant expression evaluation: preserve subobject designator when flattening a
core constant value down to an APValue.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143909 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRip the ObjCPropertyRef l-value kind out of IR-generation.
John McCall [Mon, 7 Nov 2011 03:59:57 +0000 (03:59 +0000)]
Rip the ObjCPropertyRef l-value kind out of IR-generation.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143908 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove duplicate word & unnecessary conditional.
David Blaikie [Mon, 7 Nov 2011 03:30:03 +0000 (03:30 +0000)]
Remove duplicate word & unnecessary conditional.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143907 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove tests to the appropriate directory to match the spec hierarchy.
David Blaikie [Mon, 7 Nov 2011 03:27:58 +0000 (03:27 +0000)]
Move tests to the appropriate directory to match the spec hierarchy.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143906 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAllow constexpr variables' initializers to be folded in C++11 mode. This
Richard Smith [Mon, 7 Nov 2011 03:22:51 +0000 (03:22 +0000)]
Allow constexpr variables' initializers to be folded in C++11 mode. This
partially undoes the revert in r143491, but does not introduce any new instances
of the underlying issue (which is not yet fixed) in code which does not use
the 'constexpr' keyword.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143905 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBegin lifting some of the one-off checking logic into generic helper
Chandler Carruth [Sun, 6 Nov 2011 23:39:37 +0000 (23:39 +0000)]
Begin lifting some of the one-off checking logic into generic helper
routines on the base toolchain class.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143900 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove the GCC installation detection logic down into the Generic_GCC
Chandler Carruth [Sun, 6 Nov 2011 23:39:34 +0000 (23:39 +0000)]
Move the GCC installation detection logic down into the Generic_GCC
toolchain. The logic is mostly generic already, and where possible
should be made more generic. Also, it has no impact other than to expose
a set of methods which each toolchain can then query to setup their
desired configuration. These should be available to toolchains beyond
just Linux.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143899 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove an old OpenSUSE hack that is no longer needed -- it is exactly
Chandler Carruth [Sun, 6 Nov 2011 23:10:49 +0000 (23:10 +0000)]
Remove an old OpenSUSE hack that is no longer needed -- it is exactly
the same as a directory added further down in the new logic.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143897 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove the HasMultilib check. It was essentially useless. The driver now
Chandler Carruth [Sun, 6 Nov 2011 23:09:05 +0000 (23:09 +0000)]
Remove the HasMultilib check. It was essentially useless. The driver now
looks for evidence of a multilib installation, and adds the appropriate
bits to the search paths.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143896 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCorrecting grammar.
David Blaikie [Sun, 6 Nov 2011 22:28:03 +0000 (22:28 +0000)]
Correcting grammar.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143894 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse StringRef's case transformation methods.
Benjamin Kramer [Sun, 6 Nov 2011 20:36:48 +0000 (20:36 +0000)]
Use StringRef's case transformation methods.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143889 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[arcmt] In GC, error for __strong/__weak on non-objc pointers.
Argyrios Kyrtzidis [Sun, 6 Nov 2011 18:58:23 +0000 (18:58 +0000)]
[arcmt] In GC, error for __strong/__weak on non-objc pointers.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143887 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[arcmt] In GC, clear redundant __strong's.
Argyrios Kyrtzidis [Sun, 6 Nov 2011 18:58:17 +0000 (18:58 +0000)]
[arcmt] In GC, clear redundant __strong's.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143886 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake sure when setting AttributesAsWritten of a property that they do not
Argyrios Kyrtzidis [Sun, 6 Nov 2011 18:58:12 +0000 (18:58 +0000)]
Make sure when setting AttributesAsWritten of a property that they do not
include ownership qualifiers from the type.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143885 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[arcmt] Make PropertiesRewriter::getPropertyType return the unqualified type.
Argyrios Kyrtzidis [Sun, 6 Nov 2011 18:58:07 +0000 (18:58 +0000)]
[arcmt] Make PropertiesRewriter::getPropertyType return the unqualified type.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143884 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[arcmt] Collect all the places where GC attributes __strong/__weak occur.
Argyrios Kyrtzidis [Sun, 6 Nov 2011 18:58:03 +0000 (18:58 +0000)]
[arcmt] Collect all the places where GC attributes __strong/__weak occur.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143883 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[arcmt] Refactor PropertiesRewriter to use MigrationContext.
Argyrios Kyrtzidis [Sun, 6 Nov 2011 18:57:57 +0000 (18:57 +0000)]
[arcmt] Refactor PropertiesRewriter to use MigrationContext.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143882 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agotest/Driver/crash-cleanup.c: Remove "XFAIL: mingw32" corresponding to r143876.
NAKAMURA Takumi [Sun, 6 Nov 2011 16:59:41 +0000 (16:59 +0000)]
test/Driver/crash-cleanup.c: Remove "XFAIL: mingw32" corresponding to r143876.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143877 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThe version objects need to actually store the version strings; they
Chandler Carruth [Sun, 6 Nov 2011 10:51:30 +0000 (10:51 +0000)]
The version objects need to actually store the version strings; they
aren't guaranteed to live long enough otherwise.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143875 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThrow the switch and move all Linux header search over to the GCC
Chandler Carruth [Sun, 6 Nov 2011 10:31:01 +0000 (10:31 +0000)]
Throw the switch and move all Linux header search over to the GCC
detection system that is providing the library paths and crt object
files.

This, modulo any bugs that need to be shaken out, resolves numerous bugs
with how we handle header paths. Here are a few that I know of:
- We no longer need to enumerate all GCC versions searched.
- OpenSUSE searched GCC versions in the wrong order.
- There were typos when selecting various patterns, etc.
- We aren't stating quite some many directories now.
- SysRoot didn't always work in a reasonable way.

I'm working on tests for this, but the tests are making me and Lit sad.
The real testing for this type of driver change is to try it out on
various distributions. I'll hit the common ones right away, and start
more thorough testing tomorrow after some sleep.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143874 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd a missing triple spotted by inspecting and testing of the include
Chandler Carruth [Sun, 6 Nov 2011 10:30:58 +0000 (10:30 +0000)]
Add a missing triple spotted by inspecting and testing of the include
path triples. Also order the 32-bit triples a bit more reasonably.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143873 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUnbreak gcc build.
Benjamin Kramer [Sun, 6 Nov 2011 09:50:13 +0000 (09:50 +0000)]
Unbreak gcc build.

For some reason clang and msvc accept this, but gcc doesn't.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143872 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLift the GCCVersion type into the header file and start persisting it in
Chandler Carruth [Sun, 6 Nov 2011 09:39:46 +0000 (09:39 +0000)]
Lift the GCCVersion type into the header file and start persisting it in
the detected GCC installation. This allows us to expose another aspect
of what we detected: the GCC version. This will be used shortly.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143871 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake format string 64 bit safe.
Benjamin Kramer [Sun, 6 Nov 2011 09:34:11 +0000 (09:34 +0000)]
Make format string 64 bit safe.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143870 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSwitch some of these interfaces from std::string to StringRef and Twine.
Chandler Carruth [Sun, 6 Nov 2011 09:21:54 +0000 (09:21 +0000)]
Switch some of these interfaces from std::string to StringRef and Twine.
This will facilitate further use and recombinations of them.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143869 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agotest/Driver/crash-cleanup.c: Mark as XFAIL: mingw32, for investigating.
NAKAMURA Takumi [Sun, 6 Nov 2011 09:11:35 +0000 (09:11 +0000)]
test/Driver/crash-cleanup.c: Mark as XFAIL: mingw32, for investigating.

"XFAIL: msys" has no effect, due to XFAIL inspects triplet, and msys build does not use *-msys-* but *-mingw32.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143868 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoChange the AST representation of operations on Objective-C
John McCall [Sun, 6 Nov 2011 09:01:30 +0000 (09:01 +0000)]
Change the AST representation of operations on Objective-C
property references to use a new PseudoObjectExpr
expression which pairs a syntactic form of the expression
with a set of semantic expressions implementing it.
This should significantly reduce the complexity required
elsewhere in the compiler to deal with these kinds of
expressions (e.g. IR generation's special l-value kind,
the static analyzer's Message abstraction), at the lower
cost of specifically dealing with the odd AST structure
of these expressions.  It should also greatly simplify
efforts to implement similar language features in the
future, most notably Managed C++'s properties and indexed
properties.

Most of the effort here is in dealing with the various
clients of the AST.  I've gone ahead and simplified the
ObjC rewriter's use of properties;  other clients, like
IR-gen and the static analyzer, have all the old
complexity *and* all the new complexity, at least
temporarily.  Many thanks to Ted for writing and advising
on the necessary changes to the static analyzer.

I've xfailed a small diagnostics regression in the static
analyzer at Ted's request.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143867 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTake a better approach to detecting and selecting multiarch include
Chandler Carruth [Sun, 6 Nov 2011 08:21:07 +0000 (08:21 +0000)]
Take a better approach to detecting and selecting multiarch include
directories. This way we stop at the first multiarch directory found on
the system. This achieves the real intended result of pruning
non-existent directories.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143866 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert r143864, as the problem I was actually looking at can be solved
Chandler Carruth [Sun, 6 Nov 2011 08:21:04 +0000 (08:21 +0000)]
Revert r143864, as the problem I was actually looking at can be solved
a better way. The more I think about it the more worried I am that this
hammer is simply too large. We should only be reaching out to the
filesystem when doing interesting "detection" things, not gratuitously.

Original commit message:
Start pruning down the set of flags passed to CC1 for header search.
This cleans up the CC1 invocations, and reduces the overhead there.
We're still hammering the filesystem looking for the C++ standard
libraries though.

The only reservation I have about this policy is the case of virtualized
files inside of CC1, but it's not clear what the best way to solve that
is. The Driver consistently queries the actual filesystem to make its
decisions. Changing that would be a very large undertaking. It might be
worthwhile, but it's not an immediate goal.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143865 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoStart pruning down the set of flags passed to CC1 for header search.
Chandler Carruth [Sun, 6 Nov 2011 07:31:36 +0000 (07:31 +0000)]
Start pruning down the set of flags passed to CC1 for header search.
This cleans up the CC1 invocations, and reduces the overhead there.
We're still hammering the filesystem looking for the C++ standard
libraries though.

The only reservation I have about this policy is the case of virtualized
files inside of CC1, but it's not clear what the best way to solve that
is. The Driver consistently queries the actual filesystem to make its
decisions. Changing that would be a very large undertaking. It might be
worthwhile, but it's not an immediate goal.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143864 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThis test was assuming that /usr/include was in the system header search
Chandler Carruth [Sun, 6 Nov 2011 06:59:15 +0000 (06:59 +0000)]
This test was assuming that /usr/include was in the system header search
path. That assumption should never have been true, but it was until
I fixed it. Now that its fixed, add a triple here to get correct
behavior even on Windows.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143863 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoXFAIL crash-cleanup.c only on Darwin and msys.
Peter Collingbourne [Sun, 6 Nov 2011 04:32:01 +0000 (04:32 +0000)]
XFAIL crash-cleanup.c only on Darwin and msys.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143858 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoXFAIL test until Peter has a chance to investigate.
Chad Rosier [Sun, 6 Nov 2011 03:42:36 +0000 (03:42 +0000)]
XFAIL test until Peter has a chance to investigate.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143857 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFixing -g break caused by a dead reference to a temporary introduced in r143732.
David Blaikie [Sun, 6 Nov 2011 01:53:39 +0000 (01:53 +0000)]
Fixing -g break caused by a dead reference to a temporary introduced in r143732.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143856 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd the depfile as a result file so that it is deleted if we crash,
Peter Collingbourne [Sun, 6 Nov 2011 00:40:05 +0000 (00:40 +0000)]
Add the depfile as a result file so that it is deleted if we crash,
and prevent the diagnostic preprocessor run from writing a depfile.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143846 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSwitch Lit to directly query the driver for the builtin inclue path.
Chandler Carruth [Sat, 5 Nov 2011 23:29:28 +0000 (23:29 +0000)]
Switch Lit to directly query the driver for the builtin inclue path.
Thanks to Peter for pointing out how easy this is to do. I'm now much
happier with this solution.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143842 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnhance the GCC version parsing and comparison logic to handle some more
Chandler Carruth [Sat, 5 Nov 2011 23:24:30 +0000 (23:24 +0000)]
Enhance the GCC version parsing and comparison logic to handle some more
edge cases and have better behavior. Specifically, we should actually
prefer the general '4.6' version string over the '4.6.1' string, as
'4.6.2' should be able to replace it without breaking rpaths or any
other place that these paths have been embedded. Debian-based
distributions are already using a path structure with symlinks to
achieve in-place upgrades for patch versions. Now our parsing reflects
this and we select the shorter paths instead of the longer paths.

A separate issue was that we would not parse a leading patch version
number even in the presence of a suffix. The above change makes this
more problematic as it would cause a suffix being added to make us treat
the entire thing as patch-version-agnostic, which it isn't. This changes
the logic to distinguish between '4.4.x' and 4.4.1-x', and retain that
the latter has *some* patch number information. Currently, we always
bias toward the shorter and more canonical version strings. If it
becomes important we can add more Debian like rules to produce sequences
such as '4.4.1b' > '4.4.1' > '4.4.1-rc3' > '4.4.1-rc2' > '4.4.1-pre5',
but I'm very doubtful this will ever matter or be desirable.

I've made the tests for this logic a bit more interesting, and added
some specific tests for logic that is now different.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143841 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove a pointless member. I have no idea why I made this not a local
Chandler Carruth [Sat, 5 Nov 2011 22:23:14 +0000 (22:23 +0000)]
Remove a pointless member. I have no idea why I made this not a local
variable to begin with... As I'm planning to add include root
information to this object, this would have caused confusion. It didn't
even *actually* hold the include root by the time we were done with it.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143840 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove a no-longer needed helper function. Thanks for implementing this
Chandler Carruth [Sat, 5 Nov 2011 22:23:11 +0000 (22:23 +0000)]
Remove a no-longer needed helper function. Thanks for implementing this
Michael!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143839 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove the GCC installation detection helper a member of the Linux
Chandler Carruth [Sat, 5 Nov 2011 22:07:51 +0000 (22:07 +0000)]
Move the GCC installation detection helper a member of the Linux
toolchain instead of merely using it in the constructor. This will allow
us to query it when building include paths as well as the file search
paths built in the constructor. I've lifted as little of it as I could
into the header file.

Eventually this will likely sink down into some of the Generic
toolchains and be used on more platforms, but I'm starting on Linux so
I can work out all the APIs needed there, where it is easiest to test
and we have the most pressing need.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143838 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove support for Gentoo subversion-ebuild installed llvm-gcc libstdc++
Chandler Carruth [Sat, 5 Nov 2011 21:26:18 +0000 (21:26 +0000)]
Remove support for Gentoo subversion-ebuild installed llvm-gcc libstdc++
headers. As llvm-gcc is dead, and I have no idea if this ever really
worked, I think it's time for it to go. More importantly, it makes it
harder to generalize the include search logic. If someone really wants
these to work, they can set the CPLUS_INCLUDE_PATH environment variable.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143836 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach lit to ask the Clang it is running what version string to use
Chandler Carruth [Sat, 5 Nov 2011 20:55:50 +0000 (20:55 +0000)]
Teach lit to ask the Clang it is running what version string to use
rather than presuming that it is 3.0. This is extra important as the
version should be 3.1, but CMake hasn't caught up with the times.
That'll be fixed in a separate commit.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143823 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove the Linux header searching from the Frontend to the Driver. This is
Chandler Carruth [Sat, 5 Nov 2011 20:17:13 +0000 (20:17 +0000)]
Move the Linux header searching from the Frontend to the Driver. This is
the first (and diff-noisiest) step to making Linux header searching
tremendously more principled and less brittle. Note that this step
should have essentially no functional impact. We still search the exact
same set of paths in the exact same order. The only change here is where
the code implementing such a search lives.

This has one obvious negative impact -- we now pass a ludicrous number
of flags to the CC1 layer. That should go away as I re-base this logic
on the logic to detect a GCC installation. I want to do this in two
phases so the bots can tell me if this step alone breaks something, and
so that the diffs of the refactoring make more sense.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143822 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTry to reduce the nastiness that creeps in through Windows.h a bit.
Benjamin Kramer [Sat, 5 Nov 2011 17:43:55 +0000 (17:43 +0000)]
Try to reduce the nastiness that creeps in through Windows.h a bit.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143816 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[driver] If we're only linking, don't warn about unused arguments which are
Chad Rosier [Sat, 5 Nov 2011 15:56:33 +0000 (15:56 +0000)]
[driver] If we're only linking, don't warn about unused arguments which are
obviously only used during compilation.
rdar://10386708

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143813 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCompute the path properly on different platforms. Specifically, compute
Chandler Carruth [Sat, 5 Nov 2011 10:41:42 +0000 (10:41 +0000)]
Compute the path properly on different platforms. Specifically, compute
it the exact same way that the Clang code computes this path.

Hopefully with this, the MSVC bots will actually come back to life.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143807 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoChange this test to reflect the state we are moving in. The Clang
Chandler Carruth [Sat, 5 Nov 2011 10:15:33 +0000 (10:15 +0000)]
Change this test to reflect the state we are moving in. The Clang
builtin headers are no longer going to receive the old 'implicit extern
"C" block' semantics. This hint is actually ignored by both Clang and
GCC at this point, and Clang's own builtin headers can simply be changed
if there is any issue with this. Clang should be free to include these
however it wants, and so shorter and simpler is better.

Note: *nothing* is changing about the *system* stddef.h include. That
should always have the exact same include semantics, whether with Clang
or GCC or any other compiler. Only the compiler-builtin header search
path is changing.

If anyone knows of some risk that this introduces that I've not thought
of, please chime in. So far, only Windows has switched to the Brave New
World, but others should be switching soon.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143806 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSwitch these two tests to use the Clang driver instead of CC1. They want
Chandler Carruth [Sat, 5 Nov 2011 10:15:30 +0000 (10:15 +0000)]
Switch these two tests to use the Clang driver instead of CC1. They want
to do "realistic" includes, and so need the header search logic now in
the driver. This in turn requires switching the CC1 options to the
actual driver options, and passing -Xclang where there is no analogy.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143805 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach Lit to pass the CC1 invocation the builtin include directory. This
Chandler Carruth [Sat, 5 Nov 2011 10:15:27 +0000 (10:15 +0000)]
Teach Lit to pass the CC1 invocation the builtin include directory. This
is a pretty gross hack, but I don't have any significantly cleaner ideas
for this. There are several things obviously gross about it:

1) Lit shouldn't know that Clang needs this. This really that bad, as
   Lit already knows about CC1 and other internal details.
2) This hard codes the '3.0' version number, which is pretty lame.
3) This hard codes every other aspect of the resource dir structure
   which is less lame than the version number, but still not great.

However, it should bring the MSVC tests back to life, and it should
unblock the rest of the move from Frontend to Driver, so I think it's
worth a bit of grossness that is isolated in our testing infrastructure
while we figure out the best long term approach. I have the following
ideas, some of which only solve part of the problem (and thus might need
to be combined with other ideas):

a) Create a symlink or other convenience path instead of a version
   number.
b) Run 'clang' directly in the lit.cfg, look at its resource dir, and use
   that.
c) Switch all the tests to use the driver instead of CC1.
d) Hack the frontend to synthesize builtin include directories when none
   are provided by the driver.

I don't like (d) because it feels very hackish and likely to break. We
can only solve a small part of the problem with (a). I wanted to vote
for (c), but lots of the tests in this bucket are really heavily using
internal-only flags like -verify and -triple. I'm loath to complicate
them with the full driver layer. Also, switching them to the driver adds
more than just builtin headers, but all of the rest of the system
headers!

This leaves me with (b). If others like (b), I'll switch to it, but it
felt a bit icky. Nothing concrete, and the other options look
significantly worse, but I felt icky enough that I wanted to start with
a more brain-dead patch to stop the bleeding, and gauge others' feelings
here.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143804 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix markup weirdness.
Benjamin Kramer [Sat, 5 Nov 2011 10:11:36 +0000 (10:11 +0000)]
Fix markup weirdness.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143803 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix a significant oversight in my move of MSVC includes to the driver:
Chandler Carruth [Sat, 5 Nov 2011 09:24:44 +0000 (09:24 +0000)]
Fix a significant oversight in my move of MSVC includes to the driver:
actually manage the builtin header file includes as well as the system
ones.

This one is actually debatable whether it belongs in the driver or not,
as the builtin includes are really an internal bit of implementation
goop for Clang. However, they must be included at *exactly* the right
point in the sequence of header files, which makes it essentially
impossible to have this be managed by the Frontend and the rest by the
Driver. I have terrible ideas that would "work", but I think they're
worse than putting this in the driver and making the Frontend library
even more ignorant of the environment and system on which it is being
run.

Also fix the fact that we weren't properly respecting the flags which
suppress standard system include directories.

Note that this still leaves all of the Clang tests which run CC1
directly and include builtin header files broken on Windows. I'm working
on a followup patch to address that.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143801 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoNow that we have a portable 64 bit format string use it to make this simple again.
Benjamin Kramer [Sat, 5 Nov 2011 09:02:52 +0000 (09:02 +0000)]
Now that we have a portable 64 bit format string use it to make this simple again.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143800 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd two flags to the CC1 layer that I was hoping to avoid. We need to
Chandler Carruth [Sat, 5 Nov 2011 08:30:29 +0000 (08:30 +0000)]
Add two flags to the CC1 layer that I was hoping to avoid. We need to
encode the *exact* semantics which the header search paths internally
built by the Frontend layer have had, which is both non-user-provided,
and at times adding the implicit extern "C" bit to the directory entry.

There are lots of CC1 options that are very close, but none do quite
this, and they are all already overloaded for other purposes. In some
senses this makes the command lines more clean as it clearly indicates
which flags are exclusively used to implement internal detection of
"standard" header search paths.

Lots of the implementation of this is really crufty, due to the
surrounding cruft. It doesn't seem worth investing lots of time cleaning
this up as it isn't new, and hopefully *lots* of this code will melt
away as header search inside of the frontend becomes increasingly
trivial.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143798 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTweak LookThroughStmt() in LiveVariables to properly look through alternativing Paren...
Ted Kremenek [Sat, 5 Nov 2011 07:34:28 +0000 (07:34 +0000)]
Tweak LookThroughStmt() in LiveVariables to properly look through alternativing ParenExprs and OpaqueValueExprs.  Thanks to Anna and Argiris for iterating on this function.  My original patch embarssingly didn't even pass the Clang tests.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143797 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLookThroughStmt GM release.
Argyrios Kyrtzidis [Sat, 5 Nov 2011 07:09:40 +0000 (07:09 +0000)]
LookThroughStmt GM release.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143796 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLookThroughStmt v4.6.3
Argyrios Kyrtzidis [Sat, 5 Nov 2011 07:06:54 +0000 (07:06 +0000)]
LookThroughStmt v4.6.3

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143795 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSimplify LookThroughStmt in LiveVariables.cpp
Argyrios Kyrtzidis [Sat, 5 Nov 2011 07:03:54 +0000 (07:03 +0000)]
Simplify LookThroughStmt in LiveVariables.cpp

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143794 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix vector macros to correctly check argument types. <rdar://problem/10261670>
Bob Wilson [Sat, 5 Nov 2011 06:08:06 +0000 (06:08 +0000)]
Fix vector macros to correctly check argument types. <rdar://problem/10261670>

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143792 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agotests: Remove some stray and unused -cc1 flags. These tests appear to be written
Daniel Dunbar [Sat, 5 Nov 2011 05:41:57 +0000 (05:41 +0000)]
tests: Remove some stray and unused -cc1 flags. These tests appear to be written
to go through the driver.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143791 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAnother correction to the LiveVariables commit (r143767, r143780).
Anna Zaks [Sat, 5 Nov 2011 05:20:59 +0000 (05:20 +0000)]
Another correction to the LiveVariables commit (r143767, r143780).
Make test/Analysis/misc-ps.c test pass.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143790 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[analyzer] Relax the cmp criteria for analyzer reports.
Anna Zaks [Sat, 5 Nov 2011 05:20:56 +0000 (05:20 +0000)]
[analyzer] Relax the cmp criteria for analyzer reports.

Until we find out a way to easily find out what changed by looking at the logs.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143789 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[analyzer] Add the list of available checkers to the website.
Anna Zaks [Sat, 5 Nov 2011 05:20:54 +0000 (05:20 +0000)]
[analyzer] Add the list of available checkers to the website.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143788 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[analyzer] There should be a space between "expect" and "only"
Anna Zaks [Sat, 5 Nov 2011 05:20:51 +0000 (05:20 +0000)]
[analyzer] There should be a space between "expect" and "only"

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143787 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[analyzer] Add support for testing with individual preprocessed files.
Anna Zaks [Sat, 5 Nov 2011 05:20:48 +0000 (05:20 +0000)]
[analyzer] Add support for testing with individual preprocessed files.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143786 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix infinite loop in LiveVariables due to a misplaced 'break' (it would break out of
Argyrios Kyrtzidis [Sat, 5 Nov 2011 04:03:43 +0000 (04:03 +0000)]
Fix infinite loop in LiveVariables due to a misplaced 'break' (it would break out of
switch statement, not the while loop).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143780 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[libclang] For "c-index-test -index-file", print out the names of the callbacks.
Argyrios Kyrtzidis [Sat, 5 Nov 2011 04:03:35 +0000 (04:03 +0000)]
[libclang] For "c-index-test -index-file", print out the names of the callbacks.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143779 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoOn Linux, enable the gold plugin if we are using LTO.
Peter Collingbourne [Sat, 5 Nov 2011 03:47:53 +0000 (03:47 +0000)]
On Linux, enable the gold plugin if we are using LTO.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143778 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCorrect the grammar for __builtin_astype.
Peter Collingbourne [Sat, 5 Nov 2011 03:47:48 +0000 (03:47 +0000)]
Correct the grammar for __builtin_astype.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143777 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoserialized diagnostics: include FixIt information in serialized diagnostics.
Ted Kremenek [Sat, 5 Nov 2011 03:34:23 +0000 (03:34 +0000)]
serialized diagnostics: include FixIt information in serialized diagnostics.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143776 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove definitions of SizeType and PtrDiffType to Mips32TargetInfoBase.
Akira Hatanaka [Sat, 5 Nov 2011 01:48:34 +0000 (01:48 +0000)]
Move definitions of SizeType and PtrDiffType to Mips32TargetInfoBase.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143775 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix test so that it XFAILs consistently.
Eli Friedman [Sat, 5 Nov 2011 00:49:47 +0000 (00:49 +0000)]
Fix test so that it XFAILs consistently.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143771 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix MSVC build.
Michael J. Spencer [Sat, 5 Nov 2011 00:46:46 +0000 (00:46 +0000)]
Fix MSVC build.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143770 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix one last place where we weren't writing into a string literal consistently.
Eli Friedman [Sat, 5 Nov 2011 00:41:04 +0000 (00:41 +0000)]
Fix one last place where we weren't writing into a string literal consistently.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143769 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix incorrect format string in debug output.
Eli Friedman [Sat, 5 Nov 2011 00:38:30 +0000 (00:38 +0000)]
Fix incorrect format string in debug output.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143768 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach LiveVariables to look through OpaqueValueExprs for extending Stmt liveness.
Ted Kremenek [Sat, 5 Nov 2011 00:26:53 +0000 (00:26 +0000)]
Teach LiveVariables to look through OpaqueValueExprs for extending Stmt liveness.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143767 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoPer discussion with John McCall, don't add OpaqueValueExprs to the CFG.
Ted Kremenek [Sat, 5 Nov 2011 00:10:15 +0000 (00:10 +0000)]
Per discussion with John McCall, don't add OpaqueValueExprs to the CFG.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143766 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoserialized diagnostics: pull emission of filenames into diagnostic block.
Ted Kremenek [Sat, 5 Nov 2011 00:10:11 +0000 (00:10 +0000)]
serialized diagnostics: pull emission of filenames into diagnostic block.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143765 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoserialized diagnostics: pull emission of diagnostic flag string into diagnostics...
Ted Kremenek [Sat, 5 Nov 2011 00:10:07 +0000 (00:10 +0000)]
serialized diagnostics: pull emission of diagnostic flag string into diagnostics block.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143764 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSerialized diagnostics: pull category name serialization into diagnostic blocks....
Ted Kremenek [Sat, 5 Nov 2011 00:10:04 +0000 (00:10 +0000)]
Serialized diagnostics: pull category name serialization into diagnostic blocks.  The goal is to remove BLOCK_STRINGS so that the bitcode file can potentially be streamed.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143763 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoserialized diagnostics: emit source ranges.
Ted Kremenek [Sat, 5 Nov 2011 00:10:01 +0000 (00:10 +0000)]
serialized diagnostics: emit source ranges.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143762 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove definition of record/block IDs for serialized diagnostics to public header.
Ted Kremenek [Sat, 5 Nov 2011 00:09:57 +0000 (00:09 +0000)]
Move definition of record/block IDs for serialized diagnostics to public header.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143761 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSerialized diagnostics: encode a 24 bit version number as part of the "magic number...
Ted Kremenek [Sat, 5 Nov 2011 00:09:53 +0000 (00:09 +0000)]
Serialized diagnostics: encode a 24 bit version number as part of the "magic number" of serialized diagnostics.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143760 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoserialized diagnostics: unique warning strings when emitting serialized diagnostics.
Ted Kremenek [Sat, 5 Nov 2011 00:09:50 +0000 (00:09 +0000)]
serialized diagnostics: unique warning strings when emitting serialized diagnostics.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@143759 91177308-0d34-0410-b5e6-96231b3b80d8