aggregate types in a profoundly wrong way that has to be
worked around in every call site, to getEvaluationKind,
which classifies and distinguishes between all of these
cases.
Also, normalize the API for loading and storing complexes.
I'm working on a larger patch and wanted to pull these
changes out, but it would have be annoying to detangle
them from each other.
llvm-svn: 176656
With the cursor located at "I", clang-format would not do anything to:
int a;
I
int b;
With this patch, it reduces the number of empty lines as necessary, and
removes unnecessary whitespace. It does not change/reformat "int a;" or
"int b;".
llvm-svn: 176650
rdar:13370002 [pre-RA-sched] assertion: released too many times
I tracked this down to an earlier hack that is no longer applicable
and interfered with normal scheduler logic. With the changes in
r176037, it was causing an instruction to be scheduled multiple times.
I have an external test case that I tried hard to reduce and
failed. I can't even reproduce with llc.
llvm-svn: 176636
That can usually be lowered efficiently and is common in sandybridge code.
It would be nice to do this in DAGCombiner but we can't insert arbitrary
BUILD_VECTORs this late.
Fixes PR15462.
llvm-svn: 176634
- change string "ERROR" to "FAIL" to match clang lit test results
Also, make LLDB tests work on machines that do not have svn and/or git installed
llvm-svn: 176633
For iterators where the dereference operator returns by value, LoopConvert
should use 'auto &&' in the range-based for loop expression.
If the dereference operator returns an rvalue reference, this is deemed too
strange and the for loop is not converted.
Moved test case from iterator_failing.cpp to iterator.cpp and added extra
tests.
Fixes PR15437.
Reviewer: gribozavr
llvm-svn: 176631
NOTE: You may need to run 'make clean' or 'ninja -t clean' etc!!! This
is due to really nasty bug/interactions between
CMake/configure/make/Ninja/LIT...
This commit tries to back out the support for generating test cases as
part of the build system due to the issues I brought up in post-commit
review:
1) It adds a *lot* of complexity and fragility to the build system. See
the number of commits required to try to get all the bots happy.
2) It isn't really necessary -- we can already run scripts to generate
things with the RUN lines of a test.
3) It makes the tests somewhat harder to debug as they cross between
more domains.
4) In almost all cases it isn't really needed or it can be done directly
using the preprocessor.
I should have been more proactive reviewing this, and I'm really sorry
about the churn here. =/ To help keep track of what commits are going
where, this backs out most of the non-test-changes from these revisions:
r176397
r176373
r176293
r176184
r175744
r175624
r175545
r175544
There were several trivial or cleanup changes to the lit files or other
files. Some of these looked ok, but I didn't try to tease them apart...
Edwin, if you know what to look for, please carry on with the cleanups
there, and sorry for hosing stuff here but I'm not much of a Python
person, and so I was erring on the side of cautiously backing out the
change.
I've tried to preserve the test changes everywhere I could, but review
is appreciated here in case I missed some.
I then re-wrote the tests to use the preprocessor rather than python to
expand to the various bits of code. The nicest part of this is that now
all the files are just C++ code. They edit and behave like C++ code,
etc. RUN lines with different -D flags are used to run the same test
over multiple different configurations, and includes bracketed in
special defines are used to flesh out a collection of standard interface
stubs to test interactions between pieces. These probably aren't perfect
yet, but I think its an improvement (at least in terms of build system
complexity) and will hopefully be a useful demonstration of the
technique I prefer for these types of tests.
llvm-svn: 176627
v2: update CMakeLists.txt as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 176626
v2: fix R600 regressions
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 176624
Just encode the type as target specific attribute.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 176622
string to be emitted, and two properties about the files themselves.
Use $PWD to absolut-ify the path to the coverage file. Yes, this is what GCC
does. Reverts my own r175706.
llvm-svn: 176617
into the actual gcov file.
Instead of using the bottom 4 bytes as the function identifier, use a counter.
This makes the identifier numbers stable across multiple runs.
llvm-svn: 176616
"compile_units" returns an array of all compile units in a module as a list() of lldb.SBCompileUnit objects.
"compile_unit" returns a compile unit accessor object that allows indexed access, search by full or partial path, or by regex:
(lldb) script
comp_unit = lldb.target.module['TextEdit'].compile_unit['Document.m']
comp_unit = lldb.target.module['TextEdit'].compile_unit['/path/to/Document.m']
comp_unit = lldb.target.module['TextEdit'].compile_unit[0]
comp_unit = lldb.target.module['TextEdit'].compile_unit[1]
for comp_unit in lldb.target.module['TextEdit'].compile_unit[re.compile("\.m$")]
print comp_unit
This helps do quick searches and scripting while debugging.
llvm-svn: 176613
Warn about null pointer dereference earlier when a reference to a null pointer is
passed in a call. The idea is that even though the standard might allow this, reporting
the issue earlier is better for diagnostics (the error is reported closer to the place where
the pointer was set to NULL). This also simplifies analyzer’s diagnostic logic, which has
to track “where the null came from”. As a consequence, some of our null pointer
warning suppression mechanisms started triggering more often.
TODO: Change the name of the file and class to reflect the new check.
llvm-svn: 176612
Jakub Staszak