isas and corresponding names from the underlying process
in a manner much quicker than the current approach.
The current approach accesses memory in the underlying
process with a random-access pattern as it walks across
the data structures associated with each isa. This
involves a great deal of back-and-forth with debugserver,
resulting in performance problems, especially with iOS
targets.
The new approach attempts to run an expression in the
target that collects the isas and names into two large
buffers, which LLDB then collects in one shot.
<rdar://problem/12914539>
llvm-svn: 172574
controlled by the --unwind-on-error flag, and --ignore-breakpoint which separately controls behavior when a called
function hits a breakpoint. For breakpoints, we don't unwind, we either stop, or ignore the breakpoint, which makes
more sense.
Also make both these behaviors globally settable through "settings set".
Also handle the case where a breakpoint command calls code that ends up re-hitting the breakpoint. We were recursing
and crashing. Now we just stop without calling the second command.
<rdar://problem/12986644>
<rdar://problem/9119325>
llvm-svn: 172503
Fixed an issue with the auto loading of script resources in debug info files. Any platform can add support for this, and on MacOSX we allow dSYM files to contain python modules that get automatically loaded when a dSYM file is associated with an executable or shared library.
The modifications will now:
- Let the module locate the symbol file naturally instead of using a function that only works in certain cases. This helps us to locate the script resources as long as the dSYM file can be found.
- Don't try and do any of this if the script interpreter has scripting disabled.
- Allow more than one scripting resource to be found in a symbol file by returning the list
- Load the scripting resources when a symbol file is added via the "target symbols add" command.
- Be smarter about matching the dSYM mach-o file to an existing executable in the target images by stripping extensions on the symfile basname if needed.
llvm-svn: 172275
Python OS plug-ins now fetch thread registers lazily.
Also changed SBCommandInterpreter::HandleCommand() to not take the API lock. The logic here is that from the command line you can execute a command that might result in another thread (like the private process thread) to execute python or run any code that can re-enter the public API. When this happens, a deadlock immediately occurs for things like "process launch" and "process attach".
llvm-svn: 171901
The results from Clang name lookups changed to
be ArrayRefs, so I had to change the way we
check for the presence of a result and the way
we iterate across results.
llvm-svn: 170927
When looking for the Python script to load for a given module, replace dots with underscores if there are any after stripping the extension
This means that for a module named foo.devel.xyz a file named foo_devel.py will be what we try to load
llvm-svn: 170633
for reporting class types from Objective-C runtime
class symbols. Instead, LLDB now queries the
Objective-C runtime for class types.
We have also added a (minimal) Objective-C runtime
type vendor for Objective-C runtime version 1, to
prevent regressions when calling class methods in
the V1 runtime.
Other components of this fix include:
- We search the Objective-C runtime in a few more
places.
- We enable enumeration of all members of
Objective-C classes, which Clang does in certain
circumstances.
- SBTarget::FindFirstType and SBTarget::FindTypes
now query the Objective-C runtime as needed.
- I fixed several test cases.
<rdar://problem/12885034>
llvm-svn: 170601
Update the debugserver "qProcessInfo" implementation to return the
cpu type, cpu subtype, OS and vendor information just like qHostInfo
does so lldb can create an ArchSpec based on the returned values.
Add a new GetProcessArchitecture to GDBRemoteCommunicationClient akin
to GetHostArchitecture. If the qProcessInfo packet is supported,
GetProcessArchitecture will return the cpu type / subtype of the
process -- e.g. a 32-bit user process running on a 64-bit x86_64 Mac
system.
Have ProcessGDBRemote set the Target's architecture based on the
GetProcessArchitecture when we've completed an attach/launch/connect.
llvm-svn: 170491
to report a structure with an array of size 1
at the end without accounting for that array
when reporting the struct's total size to Clang.
LLDB now coerces such an array to size 0.
<rdar://problem/12822204>
llvm-svn: 170168
equality can be strict or loose and we want code to
explicitly choose one or the other.
Also renamed the Compare function to IsEqualTo, to
avoid confusion.
<rdar://problem/12856749>
llvm-svn: 170152
- remove unused members
- add NO_PEDANTIC to selected Makefiles
- fix return values (removed NULL as needed)
- disable warning about four-char-constants
- remove unneeded const from operator*() declaration
- add missing lambda function return types
- fix printf() with no format string
- change sizeof to use a type name instead of variable name
- fix Linux ProcessMonitor.cpp to be 32/64 bit friendly
- disable warnings emitted by swig-generated C++ code
Patch by Matt Kopec!
llvm-svn: 169645
When using the same-device optimization for shared cache libraries, if
we have an invalid load address for __LINKEDIT, don't try to read
anything out of lldb's own address space. Reading it out of the remote
address space will fail gracefully if we have bad addresses but reading
it out of lldb's own address space will result in a crash.
llvm-svn: 169582
I modified the "Args::StringtoAddress(...)" function to be able to evaluate address expressions. This is now used for any command line arguments or options that takes addresses like:
memory read <addr> [<end-addr>]
memory write <addr>
breakpoint set --address <addr>
disassemble --start-address <addr> --end-address <addr>
It calls the expression parser to evaluate the address expression and will also work around the issue where the compiler doesn't like to add offsets to function pointers (which is what happens when you try to evaluate "main + 12"). So there is a temp fix in the Args::StringtoAddress() to work around this until we can get special compiler support for debug expressions with function pointers.
llvm-svn: 169556
Fixed zero sized arrays to work correctly. This will only happen once we get a clang that emits correct debug info for zero sized arrays. For now I have marked the TestStructTypes.py as an expected failure.
llvm-svn: 169465
- Removed the BitfieldMap class because it is unnecessary.
We now just track the most recently added field.
- Moved the code that calculates bitfield widths so it
can also be used to determine whether it's necessary
to insert anonymous fields.
- Simplified the anonymous field calculation code into
three cases (two of which are resolved identically).
- Beefed up the bitfield testcase.
llvm-svn: 169449
I filed this to track the fix:
<rdar://problem/12817918> Recent i386 ABI changes break the i386 test suite because stack backtracing is busted when stopped in trampolines
llvm-svn: 169409
- add new header lldb-python.h to be included before other system headers
- short term fix (eventually python dependencies must be cleaned up)
Patch by Matt Kopec!
llvm-svn: 169341
these are the *non-volatile* registers on Darwin/i386, not the
volatile registers.
Recognize the sp, pc, fp generic reg names as well.
llvm-svn: 169316
- use const char* instead of char* as needed in ObjC language runtime plugin
- use int to iterate through enum (operator++ on enum not defined)
- use initializer list instead of inline initialization of const field
llvm-svn: 169185
that kexts were newly added.
The Darwin userland dynamic loader provides lldb with a list of
newly-added or newly-removed binaries but in the kernel case we
only know that something has changed. DynamicLoaderDarwinKernel
really needs to maintain its own persistent list of kexts that
it has been notified about (most importantly, it will not detect
kext unlods) but for now we'll at least avoid re-adding an already
present kext.
<rdar://problem/12658487>, <rdar://problem/12658487>
llvm-svn: 169082
versions of UnknownAnyTy for ObjectiveC value types.
<unknown type>* makes no sense and can cause the
parser to behave very oddly.
<rdar://problem/12518999>
llvm-svn: 168844
Allow the expression parser to see more than just data symbols. We now accept any symbol that has an address. We take precautions to only accept symbols by their mangled or demangled names only if the demangled name was not synthesized. If the demangled name is synthesized, then we now mark symbols accordingly and only compare against the mangled original name.
llvm-svn: 168668
- Handle EINVAL return code from ptrace(GETSIGINFO, ...): not an error, but 'group-stop' state on Linux
- propagate SIGSTOP to inferior in above case
- this commit resolves the failure in expression_command/timeout testcase
Thanks to Sean Callanan & Matt Kopec for helping debug this problem
llvm-svn: 168523
allowed volatile registers to be returned up the stack. That leads
to unexpected/incorrect values provided to the user and we need to
avoid that.
<rdar://problem/12714247>
llvm-svn: 168123
Fixed an issue where lldb was setting breakpoints on too many methods when a partial function name with namespaces or class qualifiers was used. For example setting a breakpoint of "Foo::dealloc" was accidentally settings breakpoints on all objective C functions whose selector was "dealloc"...
llvm-svn: 168053
expressions that refer to ivars will not work because Clang
emits IR that refers to them to get the ivar offsets.
However, it is possible to search the runtime for these values.
I have added support for reading the relevant tables to the
Objective-C runtime, and extended ClangExpressionDeclMap to
query that information if and only if it doesn't find the symbols
in the binary.
Also added a testcase.
<rdar://problem/12628122>
llvm-svn: 168018
When uniquing classes against one another we can't depend on any or all of the artificial functions (default ctor, dtor, copy ctor, move ctor, etc) being in each definition. Now we treat those separately and handle those to the best of our ability.
llvm-svn: 167752
GCD queue names of threads to
ProcessGDBRemote::GetDispatchQueueNameForThread()
May need tweaking once this version is rolled out but visual
inspection looks fine.
<rdar://problem/12333100>
llvm-svn: 167667
Solving an issue where "command script import" would fail to pick the file indicated by the user as a result of something with the same name being in an earlier position in sys.path
llvm-svn: 167570
This commit does three things:
(a) introduces a new notification model for adding/removing/changing modules to a ModuleList, and applies it to the Target's ModuleList, so that we make sure to always trigger the right set of actions
whenever modules come and go in a target. Certain spots in the code still need to "manually" notify the Target for several reasons, so this is a work in progress
(b) adds a new capability to the Platforms: locating a scripting resources associated to a module. A scripting resource is a Python file that can load commands, formatters, ... and any other action
of interest corresponding to the loading of a module. At the moment, this is only implemented on Mac OS X and only for files inside .dSYM bundles - the next step is going to be letting
the frameworks themselves hold their scripting resources. Implementors of platforms for other systems are free to implement "the right thing" for their own worlds
(c) hooking up items (a) and (b) so that targets auto-load the scripting resources as the corresponding modules get loaded in a target. This has a few caveats at the moment:
- the user needs to manually add the .py file to the dSYM (soon, it will also work in the framework itself)
- if two modules with the same name show up during the lifetime of an LLDB session, the second one won't be able to load its scripting resource, but will otherwise work just fine
llvm-svn: 167569
DynamicLoaderDarwinKernel::OSKextLoadedKextSummary::LoadImageUsingMemoryModule
where it assumed that a kernel had been found in memory, when that may not be
the case when we're attaching to a device early in the boot process.
<rdar://problem/12638140>
llvm-svn: 167564
<rdar://problem/12153915> (partial fix)
Remove an assert and place an error message instead so we don't crash when we run into a type tag that we don't recognize. We will now emit a warning so that hopefully we can get a bug report that has example code that shows what we are missing.
Also fixed a case when trying to unique one type to another where we would confuse concrete instances of methods with their definitions and end up not correctly registering the types.
llvm-svn: 167557
Also fixed a bug where the Objective-C runtime
would not properly report that it found a class if
(a) it had to build the ObjCInterfaceDecl for the
class and (b) logging was enabled.
<rdar://problem/12641180>
llvm-svn: 167516
Unnamed bitfields cause struct layout problems
Synthesize unnamed bitfields when required. Most compilers don't mention unnamed bitfields in the DWARF, so we need to create them to keep clang happy with the types we create from the DWARF. We currently can't do this for ObjC since the DW_AT_bit_offset value for any direct ivars of ObjC classes as the values for these attributes are bogus. A bug has been filed on Clang to fix this, and another bug has been filed on LLDB to make sure we fix the DWARF parser once the clang fix is in by looking the the DW_AT_producer in the compile unit attributes and finding the compiler version and only enabling it for newer versions of clang.
llvm-svn: 167424
The operator== method is a synonym for IsExactMatch().
The essential difference between these two is that IsCompatibleMatch()
will say that armv7 and armv7s are compatible and return true.
IsExactMatch() will say that armv7 and armv7s are not a match.
An armv7s cpu can run either generic armv7 binaries or armv7s binaries
(the latter being tuned for it). When we're picking the slice of a
universal Mach-O file to load in an armv7s Target, we need to be able to
first look for an exact cpu subtype match (armv7s == armv7s) and failing
that, looking for a slice with a compatible architecture.
Update ObjectContainerUniversalMachO::GetObjectFile to prefer an exact
match of the cpu type, falling back to a compatible match if necessary.
<rdar://problem/12593515>
llvm-svn: 167365
LLDB now provides base class offsets (virtual and non virtual) to Clang's record layout. We previously were told this wasn't necessary, but it is when pragma pack gets involved.
llvm-svn: 167262
RegisterContextKDP_i386 was not correctly writing registers due to missing "virtual" keywords. Added the virtual keywords and made the functions pure virtual to ensure subclasses can't get away without implementing these functions.
llvm-svn: 167066
I tracked down a leak that could happen when detaching from a process where the lldb_private::Process objects would stay around forever. This was caused by a eStateDetached event that was queued up on the lldb_private::Process private state thread listener. Since process events contain shared pointers to the process, this is dangerous if they don't get consume or cleared as having the lldb_private::Process class contain a collection of things that have a shared pointer to yourself is obviously bad.
To fix this I modified the Process::Finalize() function to clear this list. The actual thing that was holding onto the ModuleSP and thus the static archive, was a stack frame. Since the process wasn't going away, it still had thread objects and they still had frames. I modified the Thread::Destroy() to clear the stack frames to ensure this further doesn't happen.
llvm-svn: 166964
so it could hold this information, and then used it to look up unfound names in the object pointer
if it exists. This gets "frame var" to work for unqualified references to ivars captured in blocks.
But the expression parser is ignoring this information still.
llvm-svn: 166860
- Only read the statically-defined isa table in the
shared cache once. Only the dynamically-constructed
isa table can change.
- Ignore the statically-defined isa table if its
version isn't what we expect.
llvm-svn: 166856
Full UnwindPlan is trying to do an impossible unwind; in that case
invalidate the Full UnwindPlan and replace it with the architecture
default unwind plan.
This is a scenario that happens occasionally with arm unwinds in
particular; the instruction analysis based full unwindplan can
mis-parse the functions and the stack walk stops prematurely. Now
we can do a simpleminded frame-chain walk to find the caller frame
and continue the unwind. It's not ideal but given the complicated
nature of analyzing the arm functions, and the lack of eh_frame
information on iOS, it is a distinct improvement and fixes some
long-standing problems with the unwinder on that platform.
This is fixing <rdar://problem/12091421>. I may re-use this
invalidate feature in the future if I can identify other cases where
the full unwindplan's unwind information is clearly incorrect.
This checkin also includes some cleanup for the volatile register
definition in the arm ABI plugin for <rdar://problem/10652166>
although work remains to be done for that bug.
llvm-svn: 166757
the function's prologue instructions so we can re-instate that prologue
if we hit an early return mid-function. Add some additional heuristics
to differentiate between prologue and epilogue instruction sequences.
This fixes the specific problem of correctly unwinding through a function
which has an epilogue one instruction after the last prologue setup
instruction has completed.
<rdar://problem/12091139>
llvm-svn: 166465
- Fixed a crash when the executable module
couldn't be found.
- Fixed a problem that made it impossible
to attach to processes in the simulator
using the SBTarget::Attach.
llvm-svn: 166355
Added commands to the KDP plug-in that allow sending raw commands through the KDP protocol. You specify a command byte and a payload as ASCII hex bytes, and the packet is created with a valid header/sequenceID/length and sent. The command responds with a raw ASCII hex string that contains all bytes in the reply including the header.
An example of sending a read register packet for the GPR on x86_64:
(lldb) process plugin packet send --command 0x07 --payload 0100000004000000
llvm-svn: 166346
plugin
dynamic-loader
macosx-kernel
(bool) disable-kext-loading
To settings can be set using:
(lldb) settings set plugin.dynamic-loader.macosx-kernel.disable-kext-loading true
I currently only hooked up the DynamicLoader plug-ins, but the code is very easy to duplicate when and if we need settings for other plug-ins.
llvm-svn: 166294
1 by the expression parser. We now correctly
report that they are of size 0. (C++ structs
are mandated to have nonzero size, and Clang marks
them as being 1 byte in size.)
<rdar://problem/12380800>
llvm-svn: 166256
Added a new setting that allows a python OS plug-in to detect threads and provide registers for memory threads. To enable this you set the setting:
settings set target.process.python-os-plugin-path lldb/examples/python/operating_system.py
Then run your program and see the extra threads.
llvm-svn: 166244
<rdar://problem/12068650>
More fixes to how we handle paths that are used to create a target.
This modification centralizes the location where and how what the user specifies gets resolved. Prior to this fix, the TargetList::CreateTarget variants took a FileSpec object which meant everyone had the opportunity to resolve the path their own way. Now both CreateTarget variants take a "const char *use_exe_path" which allows the TargetList::CreateTarget to centralize where the resolving happens and "do the right thing".
llvm-svn: 166186
to handle an addition class of early-return instructions we find in arm code:
tail-call optimziation returns where we restore the register state from the
function entry and jump directly (not branch & link) to another function --
when that other function returns, it will return to our caller.
Previously this mid-function epilogue sequence was not being correctly detected.
We would not re-instate the prologue setup instructions for the rest of the function
so unwinds would break from that point until the end of the function.
<rdar://problem/12502597>
llvm-svn: 166081
This patch fixes an issue where if lldb fails to attach to a process (ie. invalid pid) on Linux, the process monitor thread gets stuck waiting for a signal from the attach thread, which never comes due to not being signaled. It also implements StopOpThread which is used for both attach/launch cases as I'm not aware of any special handling needed for the attach case. Also, propagate 'Error' from the Detach function instead of using a bool.
llvm-svn: 166055
must push something on the stack for a function call or not. In
x86, the stack pointer is decremented when the caller's pc is saved
on the stack. In arm, the stack pointer and frame pointer don't
necessarily have to change for a function call, although most
functions need to use some stack space during their execution.
Use this information in the RegisterContextLLDB to detect invalid
unwind scenarios more accurately.
<rdar://problem/12348574>
llvm-svn: 166005
I added the ability for a process plug-in to implement custom commands. All the lldb_private::Process plug-in has to do is override:
virtual CommandObject *
GetPluginCommandObject();
This object returned should be a multi-word command that vends LLDB commands. There is a sample implementation in ProcessGDBRemote that is hollowed out. It is intended to be used for sending a custom packet, though the body of the command execute function has yet to be implemented!
llvm-svn: 165861
Dynamic type code must be efficient and fast. Now it is.
Added ObjC v1 support for getting the complete list of ISA values.
The main flow of the AppleObjCRuntime subclasses is now they must override "virtual bool UpdateISAToDescriptorMap_Impl();". This function will update the complete list of ISA values and create ClassDescriptorSP objects for each one. Now we have the complete list of valid ISA values which we can use for verification when doing dynamic typing.
Refactored a bunch of stuff so that the AppleObjCRuntime subclasses don't have to implement as many functions as they used to.
llvm-svn: 165730
Then make the Thread a Broadcaster, and get it to broadcast when the selected frame is changed (but only from the Command Line) and when Thread::ReturnFromFrame
changes the stack.
Made the Driver use this notification to print the new thread status rather than doing it in the command.
Fixed a few places where people were setting their broadcaster class by hand rather than using the static broadcaster class call.
<rdar://problem/12383087>
llvm-svn: 165640
over to simply update its cache and then look up
the descriptor in the cache. This is fine because
the cache now builds much faster (since descriptors
are minimal).
Metaclasses aren't in the cache, so I switched
the Describe method for class descriptors from using
GetClassDescriptor to manually creating an automatic
ClassDescriptorV2.
llvm-svn: 165579
The following are now derived lazily:
- The name of the class (cached);
- the instance size of the class (not cached);
The following have been removed entirely:
- Whether the class is realized. This is an
implementation detail.
- The contents of the objc_class object. That
object can be read as needed.
- Whether the class is valid. The fact that
we vended a class to begin with means it's
valid. We will only give up looking parts
of it up if they are not in the format we
expect.
llvm-svn: 165567
lighter-weight so that the cache can be populated
faster.
- I Added a ProcessWP to the runtime so I can
take it out of the individual descriptors, saving
space;
- I made the constructors for the descriptors
private so that only the runtime can invoke them;
and
- I removed the constructor that takes a ValueObject
since the logic for using a ValueObject is in the
runtime.
llvm-svn: 165549
whether we try to call an external program to load symbols unconditionally,
or if we check the user's preferences before calling it.
ProcessMachCore now sets CanJIT to false - we can't execute code in a core file.
DynamicLoaderDarwinKernel::OSKextLoadedKextSummary::LoadImageUsingMemoryModule changed
to load the kernel from an on-disk file if at all possible.
Don't load the kext binaries out of memory from the remote systems - their linkedit doesn't
seem to be in a good state and we'll error out down in SymbolVendorMacOSX if we try to use
the in-memory images.
Call Symbols::DownloadObjectAndSymbolFile to get the kext/kernel binary -- the external
program may be able to give us a file path on the local filesystem instead of reading
the binary / dSYM over a network drive every time. Fall back to calling
Target::GetSharedModule() like before if DownloadObjectAndSymbolFile fails.
llvm-svn: 165471
if we have a kernel binary, set the target's architecture to match.
Include the target's architecture in the ModuleSpec when we're searching for the
kext binaries on the local system -- otherwise we won't get a specific slice of
a fat file picked out for us and we won't use the returned Module correctly.
Remove the redundant attempt to find a file on the local filesystem from this method.
In ProcessGDBRemote::CheckForKernel(), if we have a kernel binary in memory, mark
the canJIT as false. There is no jitting code in kernel debug sessions.
llvm-svn: 165357
enabled after we'd found a few bugs that were caused by shadowed
local variables; the most important issue this turned up was
a common mistake of trying to obtain a mutex lock for the scope
of a code block by doing
Mutex::Locker(m_map_mutex);
This doesn't assign the lock object to a local variable; it is
a temporary that has its dtor called immediately. Instead,
Mutex::Locker locker(m_map_mutex);
does what is intended. For some reason -Wshadow happened to
highlight these as shadowed variables.
I also fixed a few obivous and easy shadowed variable issues
across the code base but there are a couple dozen more that
should be fixed when someone has a free minute.
<rdar://problem/12437585>
llvm-svn: 165269
Sean Callanan