This patch adds ThreadSanitizer support into LLDB:
- Adding a new InstrumentationRuntime plugin, ThreadSanitizerRuntime, in the same way ASan is implemented.
- A breakpoint stops in `__tsan_on_report`, then we extract all sorts of information by evaluating an expression. We then populate this into StopReasonExtendedInfo.
- SBThread gets a new API, SBThread::GetStopReasonExtendedBacktraces(), which returns TSan’s backtraces in the form of regular SBThreads. Non-TSan stop reasons return an empty collection.
- Added some test cases.
Reviewed by Greg Clayton.
llvm-svn: 264162
We are using hardlinks instead of symlinks, and we attempted to
have some logic where we don't re-create the link if the target
file already exists. This logic is faulty, however, when you
manually delete the source file (e.g. liblldb.dll) and then rebuild
lldb so that a brand new liblldb.dll gets written. Now the two files
have different inodes, but the target exists, so we would not remake
the link and the target would become stale.
We fix this by only doing the optimization if they are really the
exact same file (by comparing inode numbers), and if they are not
the same file but the target exists, we delete it and re-create
the link.
llvm-svn: 263844
The swig typemaps had some magic for output File *'s on OS X that made:
SBDebugger.GetOutputFileHandle()
actually work. That was protected by a "#ifdef __MACOSX__", but the corresponding define
got lost going from the Darwin shell scripts to the python scripts for running
swig, so the code was elided. I need to pass the define to SWIG, but only when
targetting Darwin.
So I added a target-platform argument to prepare_bindings, and if that
is Darwin, I pass -D__APPLE__ to swig, and that activates this code again, and
GetOutputFileHandle works again. Note, I only pass that argument for the Xcode
build. I'm sure it is possible to do that for cmake, but my cmake-foo is weak.
I should have been able to write a test for this by creating a debugger, setting the
output file handle to something file, writing to it, getting the output file handle
and reading it. But SetOutputFileHandle doesn't seem to work from Python, so I'd
have to write a pexpect test to test this, which I'd rather not do.
llvm-svn: 263183
That way you can set offset breakpoints that will move as the function they are
contained in moves (which address breakpoints can't do...)
I don't align the new address to instruction boundaries yet, so you have to get
this right yourself for now.
<rdar://problem/13365575>
llvm-svn: 263049
Summary:
This makes cloning (and therefore the whole build) faster.
The checkout step goes from ~4m to ~30s on my host.
Reviewers: tfiala
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D17425
llvm-svn: 262513
working directory by default -- a typical security problem that we
need to be more conservative about.
It adds a new target setting, target.load-cwd-lldbinit which may
be true (always read $cwd/.lldbinit), false (never read $cwd/.lldbinit)
or warn (warn if there is a $cwd/.lldbinit and don't read it). The
default is set to warn. If this is met with unhappiness, we can look
at changing the default to true (to match current behavior) on a
different platform.
This does not affect reading of ~/.lldbinit - that will still be read,
as before. If you run lldb in your home directory, it will not warn
about the presence of a .lldbinit file there.
I had to add two SB API - SBHostOS::GetUserHomeDirectory and
SBFileSpec::AppendPathComponent - for the lldb driver code to be
able to get the home directory path in an OS neutral manner.
The warning text is
There is a .lldbinit file in the current directory which is not being read.
To silence this warning without sourcing in the local .lldbinit,
add the following to the lldbinit file in your home directory:
settings set target.load-cwd-lldbinit false
To allow lldb to source .lldbinit files in the current working directory,
set the value of this variable to true. Only do so if you understand and
accept the security risk.
<rdar://problem/24199163>
llvm-svn: 261280
Summary:
This does not yet give us a clean testsuite run but it does help with:
1. Actually building on linux
2. Run the testsuite with over 70% tests passing on linux.
Reviewers: tfiala, labath, zturner
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D17182
llvm-svn: 260721
The explicit APIs on SBValue obviously remain if one wants to be explicit in intent, or override this guess, but since __int__() has to pick one, an educated guess is definitely better than than always going to signed regardless
Fixes rdar://24556976
llvm-svn: 260349
This change restores the Xcode build to working after Makefile support
was stripped from LLVM and clang recently.
With this change, the Xcode build now requires cmake (2.8.12.2+).
The cmake must either be on the path that Xcode sees, or it must
exist in one of the following locations:
* /usr/local/bin/cmake
* /opt/local/bin/cmake
* $HOME/bin/cmake
If the ninja build tool is present on the path, it will be used.
If not, ninja will be cloned (via git), bootstrap-built, and
used for the llvm/clang build.
LLDB now requires a minimum deployment target of OS X 10.9. Prior
to this, it was 10.8. The llvm/clang cmake build will not run
with Xcode 7.2 or Xcode 7.3 beta's compiler with the minimum
deployment target set to anything lower than 10.9. This is
related to #include <atomic>.
When llvm or clang source code does not exist in the lldb tree,
it will be cloned via git using http://llvm.org/git/{project}.git.
Previously it used SVN. If this causes any heartache, we can
make this smarter, autodetect an embedded svn and use svn instead.
(And/or use SVN if a git command is not available).
This change also fixes an lldb-mi linkage failure (needed
libncurses) as exposed by one of the LLVM libs.
llvm-svn: 259027
SUMMARY:
Get the load address for the address given by symbol and function.
Earlier, this was done for function only, this patch does it for symbol too.
This patch also adds TestAvoidBreakpointInDelaySlot.py to test this change.
Reviewers: clayborg
Subscribers: labath, zturner, mohit.bhakkad, sagar, jaydeep, lldb-commits
Differential Revision: http://reviews.llvm.org/D16049
llvm-svn: 258919
SBProcess::ReadMemory and other related functions such as
WriteMemory are returning Python string() objects. This means
that in Python 3 that are returning Unicode objects. In reality
they should be returning bytes objects which is the same as a string
in Python 2, but different in Python 3. This patch updates the
generated SWIG code to return Python bytes objects for all
memory related functions.
One quirk of this patch is that the C++ signature of ReadCStringFromMemory
has it writing c-string data into a void*. This confuses our swig
typemaps which expect that a void* means byte data. So I hacked up
a custom typemap which maps this specific function to treat the
void* as string data instead of byte data.
llvm-svn: 258743
This needs to be able to handle bytes, strings, and bytearray objects.
In Python 2 this was easy because bytes and strings are the same thing,
but in Python 3 the 2 cases need to be handled separately. So as not
to mix raw Python C API code with PythonDataObjects code, I've also
introduced a PythonByteArray class to PythonDataObjects to make the
paradigm used here consistent.
llvm-svn: 258741
We already have char** typemaps which were near copy-pastes of
the const char** versions. This way we have only one version that
works for both.
llvm-svn: 257670
There were a number of problems preventing this from working:
1. The SWIG typemaps for converting Python lists to and from C++
arrays were not updated for Python 3. So they were doing things
like PyString_Check instead of using the PythonString from
PythonDataObjects.
2. ProcessLauncherWindows was ignoring the environment completely.
So any test that involved launching an inferior with any kind
of environment variable would have failed.
3. The test itself was using process.GetSTDOUT(), which isn't
implemented on Windows. So this was changed to save the
value of the environment variable in a local variable and
have the debugger look at the value of the variable.
llvm-svn: 257669
* lldb::tid_t was being converted incorrectly, so this is updated to use
PythonInteger instead of manual Python Native API calls.
* OSPlugin_RegisterContextData was assuming that the result of
get_register_data was a string, when in fact it is a bytes. So this
method is updated to use PythonBytes to do the work.
llvm-svn: 257398
Summary: If six.py is simlink'd, an installation won't be able to find it unless it has access to the source tree that lldb was built from.
Reviewers: zturner
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D15422
llvm-svn: 255340
We were trying to be super smart and find all the supported language
bindings. This led to us scanning the directory and treating all
subdirectories as language binding directories. This makes it
hard to add unrelated code in this folder.
Besides, we only support one at the moment - Python. And when new
ones are added it will be trivial to just add their names to a list.
So this patch gets stupider about how to look for language binding
subfolders. Just put them in a list, and use the list.
llvm-svn: 254078
This script really should not be assuming every subdirectory is
a language directory for swig generation. Using a hack to get
this working for now, but this should be solved once this script
is re-written similar to how prepare_bindings was.
llvm-svn: 254037
With this patch, the client will package up all the required
inputs into a compressed zip file, establish a connection to the
server, send the input to the server, and wait for the server to
send a response (in this case the response is just echoed back to
the client).
This gets the network communication in place, and in a subsequent
patch I will follow up with the code that actually runs swig on
the server and sends back the output instead of echoing back the
input.
llvm-svn: 254023
This version supports local generation only. It's intentionally
stupid, and does not support any kind of dependency checking.
If you run the script, it's going to call SWIG. While this is
a slow process, we are going to combine the use of the swig bot
with checked in static bindings, meaning that it won't be terribly
important to only regenerate the bindings when the input files
have actually changed.
A side benefit of this is that the implementation is drastically
simpler.
This is all experimental at the moment, but it duplicates a lot
of the logic currently found in prepare_bindings.py. There was
not a good way to reuse some of the logic without invasive changes
on that script, and since this script is still experimental, it
makes sense to just copy them over, and if / when this becomes
more mature, we can get rid of the other ones.
llvm-svn: 254022
This patch fixes two issues:
1) Popen needs to be used with universal_newlines=True by default.
This elicits automatic decoding from bytes -> string in Py3,
and has no negative effects in other Py versions.
2) The swig typemaps for converting between string and (char*, int)
did not work correctly when the length of the string was 0,
indicating an error. In this case we would try to construct a
string from uninitialized data.
3) Ironically, the bug mentioned in #2 led to a test passing on
Windows that was actually broken, because the test was written
such that the assertion was never even getting checked, so it
passed by default. So we additionally fix this test to also
fail if the method errors. By fixing this test it's now broken
on Windows, so we also xfail it.
llvm-svn: 253487
This change does not introduce static bindings. It is simply using
the pylinted cleaned up code in prepare_bindings.py.
If this breaks anyting, I'll revert immediately and figure out what
needs to be addressed. I'm looking to wrap up
the cleanup aspect of the code change (pylinted, removal of code that
implements existing python stdlib code, fixes for Xcode adoption, etc.).
llvm-svn: 253478
Added a new flag, --allow-static-binding. When specified,
if (and only if) the swig binary cannot be found, then the
LLDBWrapPython.cpp and lldb.py from the
scripts/Python/{static-binding-dir} are copied into the place where
swig would have generated them.
{static-binding-dir} defaults to static-binding, and can be
overridden with the --static-binding-dir command line argument.
The static bindings checked in are from r253424.
llvm-svn: 253448
This is only used by Xcode at the moment. It replaces the
buildSwigWrapperClasses.py and related per-script-language
scripts. It also fixes a couple bugs in those w/r/t Xcode
usage:
* the presence of the GCC_PREPROCESSOR_DEFINITIONS env var
should not be short-circuiting generation of the language
binding; rather, only if LLDB_DISABLE_PYTHON is present
within that environment variable.
* some logic around what to do when building in "non-Makefile"
mode. I've switched the handling of that to be on a
"--framework" flag - if specified, we build an OS X-style
framework; otherwise, we go with non.
Putting this up now only attached to the Xcode build so
others can look at it but not be affected by it yet.
After this, I'll tackle the finalizer, along with trying
it locally on Linux.
llvm-svn: 253317
breakpoint as "file address" so that the address breakpoint will track that
module even if it gets loaded in a different place. Also fixed the Address
breakpoint resolver so that it handles this tracking correctly.
llvm-svn: 253308
Python 3 has lots of new debug asserts, and some of these were
firing on PythonFile. Specifically related to handling of invalid
files.
llvm-svn: 253261
This finishes the effort to port python-wrapper.swig code over to
using PythonDataObjects.
Also included in this patch is the removal of `PyCallable` from
`python-wrapper.swig`, as it is no longer used after having been
replaced by `PythonCallable` everywhere.
There might be additional cleanup as followup patches, but it should
be all fairly simple and minor.
llvm-svn: 252939
PyCallable is a class that exists solely within the swig wrapper
code. PythonCallable is a more generic implementation of the same
idea that can be used by any Python-related interop code, and lives
in PythonDataObjects.h
The CL is mostly mechanical, and it doesn't cover every possible
user of PyCallable, because I want to minimize the impact of this
change (as well as making it easier to figure out what went wrong
in case this causes a failure). I plan to finish up the rest of
the changes in a subsequent patch, culminating in the removal of
PyCallable entirely.
llvm-svn: 252906
This had been relegated to a simple forwarding function, so just
delete it in preparation of migrating all of these functions out
of python-wrapper.swig.
llvm-svn: 252803
This only begins to port python-wrapper.swig over. Since this
code can be pretty hairy, I plan to do this incrementally over a
series of patches, each time removing or converting more code
over to the PythonDataObjects code.
llvm-svn: 252788
Fixed a crash that would happen if you tried to get the name of a constructor or destructor by calling "getDeclName()" instead of calling getName() (which would assert and crash).
Added the ability to get function arguments names from SBFunction.
llvm-svn: 252622
Relying on manual Python C API calls is error prone, especially
when trying to maintain compatibility with Python 2 and Python 3.
This patch additionally fixes what appears to be a potentially
serious memory leak, in that were were incref'ing two values
returned from the session dictionary but never decref'ing them.
There was a comment indicating that it was intentional, but the
reasoning was, I believe, faulty and it resulted in a legitimate
memory leak.
Switching everything to PythonObject based classes solves both
the compatibility issues as well as the resource leak issues.
llvm-svn: 252536
For language that support such a thing, this API allows to ask whether a type is anonymous (i.e. has been given no name)
Comes with test case
llvm-svn: 252390
Python has a complicated mechanism of checking an objects truthity.
This involves a number of steps, which end with calling two private
methods on an object (if they are implemented). In Python 2 these
two methods are `__nonzero__` and `__len__`, and in Python 3 they
are `__bool__` and `__len__`. Because we *also* define a __len__
method for certain iterable types, this was triggering a situation
in Python 3 where `__nonzero__` wasn't defined, so it was calling
`__len__`, which was returning 0 (for example an SBDebugger with
no targets), and as a result the truthosity was determined to be
False.
We fix this by correctly using ` __bool__` for Python 3, and leave
the behavior under Python 2 unchanged.
Note that this fix is only implemented in the SWIG generation
python script, and not the SWIG generation shell script. Someone
more familiar than me with shell scripts will need to fix them
to support this for Python 3 if desired.
llvm-svn: 252382
instance:
break set -l c++ -r Name
will only break on C++ symbols that match Name, not ObjC or plain C symbols. This also works
for "break set -n" and there are SB API's to pass this as well.
llvm-svn: 252356
Summary:
Code that tried to find swig and then split the path into
a separate path and filename is being removed. The invoking
build system always provides the location of swig and we
don't need to split it into 2 pieces only to recombine it
a short time later.
Reviewers: zturner, domipheus
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D14415
llvm-svn: 252330
Summary:
This does a broad first pass on cleaning up a lot of the noise when
using pylint on these scripts. It mostly addresses issues of:
* Mixed tabs and spaces.
* Trailing whitespace.
* Semicolons where they aren't needed.
* Incorrect whitespace around () and [].
* Superfluous parentheses.
There will be subsequent patches with further changes that build
upon these.
Reviewers: zturner, domipheus
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D14375
llvm-svn: 252244
This reverts commit e59c95ca936f5a0a8abb987b8605fd8bf82b03b6.
This was a mistake on my part. The real problem was with my
environment. I was using a release interpreter to try to load
my debug extension module. I noticed this after I finally managed
to get into my extension module's init method, and then it segfaulted
with heap errors due to mismatched CRT (debug vs. release)
llvm-svn: 252030
In Python 2, a debug extension module required an _d suffix, so
for example the extension module `_lldb` would be backed by the file
`_lldb_d.pyd` if built in debug mode, and `_lldb.pyd` if built in
release mode. In Python 2, although undocumented, this seems to
no longer be the case, and even for a debug extension module, the
interpreter will only look for the `_lldb.pyd` name.
llvm-svn: 252026
This has apparently been broken since June, but only on non-Windows.
Perhaps nobody noticed it because if the symlink is already there
it won't try to re-create it, and nobody ever tried doing a clean
build.
In any case, I will let the original author attempt to fix this if
he is still interested. the problem is that in the normal case
of not setting BUILD_SHARED_LIBS and simply running ninja, it would
link _lldb.so to a non-existent location, creating a dangling
symlink.
llvm-svn: 251840
Summary:
Along with this, support for an optional argument to the "num_children"
method of a Python synthetic child provider has also been added. These have
been added with the following use case in mind:
Synthetic child providers currently have a method "has_children" and
"num_children". While the former is good enough to know if there are
children, it does not give any insight into how many children there are.
Though the latter serves this purpose, calculating the number for children
of a data structure could be an O(N) operation if the data structure has N
children. The new method added in this change provide a middle ground.
One can call GetNumChildren(K) to know if a child exists at an index K
which can be as large as the callers tolerance can be. If the caller wants
to know about children beyond K, it can make an other call with 2K. If the
synthetic child provider maintains state about it counting till K
previosly, then the next call is only an O(K) operation. Infact, all
calls made progressively with steps of K will be O(K) operations.
Reviewers: vharron, clayborg, granata.enrico
Subscribers: labath, lldb-commits
Differential Revision: http://reviews.llvm.org/D13778
llvm-svn: 250930
This makes LLDB launch and create a REPL, specifying no target so that the REPL
can create one for itself. Also added the "--repl-language" option, which
specifies the language to use. Plumbed the relevant arguments and errors
through the REPL creation mechanism.
llvm-svn: 250773
Python requires that Python.h is included before any std header. Not doing so
results in conflicts with standards macros such as `_XOPEN_SOURCE`. NFC.
llvm-svn: 250673
Using the Python native C API is non-portable across Python versions,
so this patch changes them to use the `PythonFile` class which hides
the version specific differences behind a single interface.
llvm-svn: 250525
a few days now where compiler-rt gets an error when trying
to run its install step (related to not being able to find
an ios version of a dylib), breaking the lldb build. I
don't know if I'm the only one seeing this or if everyone has
been doing the same hack I've been doing - removing the
compiler-rt project from the default checkout.
It's only used for the ASAN test case. So I'm temporarily
checking in my hackaround of not checking out compiler-rt
by default, I'll try to get back and look at what's actually
happening in the compiler-rt install step that is causing
the problems when built as a part of lldb.
llvm-svn: 250487
Added the ability to specify if an attach by name should be synchronous or not in SBAttachInfo and ProcessAttachInfo.
<rdar://problem/22821480>
llvm-svn: 249361
Kuba Brecka