Fixed the test case for "test/functionalities/exec/TestExec.py" on Darwin.
The issue was breakpoints were persisting and causing problems. When we exec, we need to clear out the process and target and start fresh with nothing and let the breakpoints populate themselves again. This patch correctly clears out the breakpoints and also flushes the process so that the objects (process/thread/frame) give out valid information.
llvm-svn: 194106
GetThreadOriginExtendedBacktraceTypeAtIndex methods to
SBProcess.
Add documentation for the GetQueueName and GetQueueID methods
to SBThread.
<rdar://problem/15314369>
llvm-svn: 194063
Instead of looking up registers by name, we use the generic ID when we can.
Also added code that creates an extra frame when running expressions by pushing the current PC and FP and then hooking up the FP backchain. This code is "#if 0" out for now until we can pair it with unwinder fixes.
llvm-svn: 194035
Cleaned up ClangUserExpression::Evaluate() to have only one variant that takes a "const EvaluateExpressionOptions& options" instead of taking many arguments.
The "--debug" option is designed to allow you to debug your expression by stopping at the first instruction (it enables --ignore-breakpoints=true and --unwind-on-error=false) and allowing you to step through your JIT code. It needs to be more integrated with the thread plan, so I am checking this in so Jim Ingham can make it happen.
llvm-svn: 194009
at some point in the past. We may have nothing more than a pc value
for this type of stack frame -- hopefully we'll have a pc and a
stop_id so we can track module loads and unloads over time and
symbolicate the pc at the correct point in time.
Also add a flag to indicate if the CFA for the frame is available
(a bit different from a CFA of LLDB_INVALID_ADDRESS) and also an
overall setting to indicate whether this is a history stack frame
or not. A history stack frame may not have a CFA, it may not have
a register context, it may not have variables, it may not have a
frame pointer or a stack pointer.
<rdar://problem/15314068>
llvm-svn: 193987
pure virtual base class and made StackFrame a subclass of that. As
I started to build on top of that arrangement today, I found that it
wasn't working out like I intended. Instead I'll try sticking with
the single StackFrame class -- there's too much code duplication to
make a more complicated class hierarchy sensible I think.
llvm-svn: 193983
defines a protocol that all subclasses will implement. StackFrame
is currently the only subclass and the methods that Frame vends are
nearly identical to StackFrame's old methods.
Update all callers to use Frame*/Frame& instead of pointers to
StackFrames.
This is almost entirely a mechanical change that touches a lot of
the code base so I'm committing it alone. No new functionality is
added with this patch, no new subclasses of Frame exist yet.
I'll probably need to tweak some of the separation, possibly moving
some of StackFrame's methods up in to Frame, but this is a good
starting point.
<rdar://problem/15314068>
llvm-svn: 193907
This commit reimplements the TypeImpl class (the class that backs SBType) in terms of a static,dynamic type pair
This is useful for those cases when the dynamic type of an ObjC variable can only be obtained in terms of an "hollow" type with no ivars
In that case, we could either go with the static type (+iVar information) or with the dynamic type (+inheritance chain)
With the new TypeImpl implementation, we try to combine these two sources of information in order to extract as much information as possible
This should improve the functionality of tools that are using the SBType API to do extensive dynamic type inspection
llvm-svn: 193564
Introduce a new boolean setting enable-auto-oneliner
This setting if set to false will force LLDB to not use the new compact one-line display
By default, one-line mode stays on, at least until we can be confident it works.
But now if it seriously impedes your workflow while it evolves/it works wonders but you still hate it, there's a way to turn it off
llvm-svn: 193450
Fixed an issue with reexported symbols on MacOSX by adding support for symbols re-exporting symbols. There is now a new symbol type eSymbolTypeReExported which contains a new name for the re-exported symbol and the new shared library. These symbols are only used when a symbol is re-exported as a symbol under a different name.
Modified the expression parser to be able to deal with finding the re-exported symbols and track down the actual symbol it refers to.
llvm-svn: 193101
Removing Host/Atomic.h
This header file was not being copied as part of our public API headers and this in turn was causing any plugin to link against LLDB.framework, since SharingPtr.h depends on it
Out of several possible options to fix this issue, the cleanest one is to revert LLDB to use std::atomic<>, as we are a C++11 project and should take advantage of it
The original rationale for going from std::atomic to Host/Atomic.h was that MSVC++ fails to link in CLR mode when std::atomic is used
This is a very Visual Studio/.net specific issue, which hopefully will be fixed
Until them, to allow Windows development to proceed, we are going with a targeted solution where we #ifdef include the Windows specific calls, and let everyone else use the
proper atomic support, as should be
If there is an unavoidable need for a LLDB-specific atomic header, the right way to go at it would be to make an API/lldb-atomic.h header and #ifdef the Windows dependency there
The FormatManager should not need to conditionalize use of std::atomic<>, as other parts of the LLDB internals are successfully using atomic (Address and IRExecutionUnit), so this
Win-specific hack is limited to SharingPtr
llvm-svn: 192993
queue name out of ProcessGDBRemote and in to the Platform
plugin, specifically PlatformDarwin.
Also add a Platform method to translate a dispatch_quaddr
to a QueueID, and a Thread::GetQueueID().
I'll add an SBThread::GetQueueID() next.
llvm-svn: 192949
To make this work this patch extends LLDB to:
- Explicitly track the link_map address for each module. This is effectively the module handle, not sure why it wasn't already being stored off anywhere. As an extension later, it would be nice if someone were to add support for printing this as part of the modules list.
- Allow reading the per-thread data pointer via ptrace. I have added support for Linux here. I'll be happy to add support for FreeBSD once this is reviewed. OS X does not appear to have __thread variables, so maybe we don't need it there. Windows support should eventually be workable along the same lines.
- Make DWARF expressions track which module they originated from.
- Add support for the DW_OP_GNU_push_tls_address DWARF opcode, as generated by gcc and recent versions of clang. Earlier versions of clang (such as 3.2, which is default on Ubuntu right now) do not generate TLS debug info correctly so can not be supported here.
- Understand the format of the pthread DTV block. This is where it gets tricky. We have three basic options here:
1) Call "dlinfo" or "__tls_get_addr" on the inferior and ask it directly. However this won't work on core dumps, and generally speaking it's not a good idea for the debugger to call functions itself, as it has the potential to not work depending on the state of the target.
2) Use libthread_db. This is what GDB does. However this option requires having a version of libthread_db on the host cross-compiled for each potential target. This places a large burden on the user, and would make it very hard to cross-debug from Windows to Linux, for example. Trying to build a library intended exclusively for one OS on a different one is not pleasant. GDB sidesteps the problem and asks the user to figure it out.
3) Parse the DTV structure ourselves. On initial inspection this seems to be a bad option, as the DTV structure (the format used by the runtime to manage TLS data) is not in fact a kernel data structure, it is implemented entirely in useerland in libc. Therefore the layout of it's fields are version and OS dependent, and are not standardized.
However, it turns out not to be such a problem. All OSes use basically the same algorithm (a per-module lookup table) as detailed in Ulrich Drepper's TLS ELF ABI document, so we can easily write code to decode it ourselves. The only question therefore is the exact field layouts required. Happily, the implementors of libpthread expose the structure of the DTV via metadata exported as symbols from the .so itself, designed exactly for this kind of thing. So this patch simply reads that metadata in, and re-implements libthread_db's algorithm itself. We thereby get cross-platform TLS lookup without either requiring third-party libraries, while still being independent of the version of libpthread being used.
Test case included.
llvm-svn: 192922
Extend DummySyntheticProvider to actually use debug-info vended children as the source of information
Make Python synthetic children either be valid, or fallback to the dummy, like their C++ counterparts
This allows LLDB to actually stop bailing out upon encountering an invalid synthetic children provider front-end, and still displaying the non synthetized ivar info
llvm-svn: 192741
When debugging with the GDB remote in LLDB, LLDB uses special packets to discover the
registers on the remote server. When those packets aren't supported, LLDB doesn't
know what the registers look like. This checkin implements a setting that can be used
to specify a python file that contains the registers definitions. The setting is:
(lldb) settings set plugin.process.gdb-remote.target-definition-file /path/to/module.py
Inside module there should be a function:
def get_dynamic_setting(target, setting_name):
This dynamic setting function is handed the "target" which is a SBTarget, and the
"setting_name", which is the name of the dynamic setting to retrieve. For the GDB
remote target definition the setting name is 'gdb-server-target-definition'. The
return value is a dictionary that follows the same format as the OperatingSystem
plugins follow. I have checked in an example file that implements the x86_64 GDB
register set for people to see:
examples/python/x86_64_target_definition.py
This allows LLDB to debug to any archticture that is support and allows users to
define the registers contexts when the discovery packets (qRegisterInfo, qHostInfo)
are not supported by the remote GDB server.
A few benefits of doing this in Python:
1 - The dynamic register context was already supported in the OperatingSystem plug-in
2 - Register contexts can use all of the LLDB enumerations and definitions for things
like lldb::Format, lldb::Encoding, generic register numbers, invalid registers
numbers, etc.
3 - The code that generates the register context can use the program to calculate the
register context contents (like offsets, register numbers, and more)
4 - True dynamic detection could be used where variables and types could be read from
the target program itself in order to determine which registers are available since
the target is passed into the python function.
This is designed to be used instead of XML since it is more dynamic and code flow and
functions can be used to make the dictionary.
llvm-svn: 192646
This is implemented by means of a get_dynamic_setting(target, setting_name) function vended by the Python module, which can respond to arbitrary string names with dynamically constructed
settings objects (most likely, some of those that PythonDataObjects supports) for LLDB to parse
This needs to be hooked up to the debugger via some setting to allow users to specify which module will vend the information they want to supply
llvm-svn: 192628
Fixed Module::ResolveSymbolContextForAddress() to be able to also look in the SymbolVendor's SymbolFile's ObjectFile for a more meaningful symbol when a symbol lookup finds a synthetic symbol from the main object file. This will help lookups on MacOSX as the main executable might be stripped, but the dSYM file always has a full symbol table.
llvm-svn: 192510
Added a way to set hardware breakpoints from the "breakpoint set" command with the new "--hardware" option. Hardware breakpoints are not a request, they currently are a requirement. So when breakpoints are specified as hardware breakpoints, they might fail to be set when they are able to be resolved and should be used sparingly. This is currently hooked up for GDB remote debugging.
Linux and FreeBSD should quickly enable this feature if possible, or return an error for any breakpoints that are hardware breakpoint sites in the "virtual Error Process::EnableBreakpointSite (BreakpointSite *bp_site);" function.
llvm-svn: 192491
Just pass a Target* into ObjectFileELF::GetImageInfoAddress so that
it can do the extra dereference necessary on MIPS, instead of passing
a flag back to the caller.
Review: http://llvm-reviews.chandlerc.com/D1899
llvm-svn: 192469
MIPS's .dyanamic section is read-only. Instead of using DT_DEBUG for
the pointer to dyld information it uses a separate tag DT_MIPS_RLD_MAP
which points to storage in the read-write .rld_map section, which in
turn points to the dyld information.
Review: http://llvm-reviews.chandlerc.com/D1890
llvm-svn: 192408
Based on the POSIX x86_64 register context. This is sufficient for opening
a mips64 (big endian) core file. Subsequent changes will connect the
disassembler, dynamic loader support, ABI, etc.
Review: http://llvm-reviews.chandlerc.com/D1873
llvm-svn: 192335
ObjectFile::CopyData is used to copy a block of target memory to the
caller's buffer (e.g. for "memory read"). This should be a straight
memcpy, and not byte-swapped if the target and host have different
endianness.
Add a new DataExtractor::CopyData() method that performs this straight
copy and use it in ObjectFile::CopyData().
llvm-svn: 192323
- By default, the above function will wait for at least one event
- Set wait_always=false to make the function return immediately if the process is already stopped
llvm-svn: 192301
Implement SBTarget::CreateValueFromAddress() with a behavior equivalent to SBValue::CreateValueFromAddress()
(but without the need to grab an SBValue first just as a starting point to make up another SBValue out of whole cloth)
llvm-svn: 192239
Formats (as in "type format") are now included in categories
The only bit missing is caching formats along with synthetic children and summaries, which might be now desirable
llvm-svn: 192217
This radar extends the notion of one-liner summaries to automagically apply in a few interesting cases
More specifically, this checkin changes the printout of ValueObjects to print on one-line (as if type summary add -c had been applied) iff:
this ValueObject does not have a summary
its children have no synthetic children
its children are not a non-empty base class without a summary
its children do not have a summary that asks for children to show up
the aggregate length of all the names of all the children is <= 50 characters
you did not ask to see the types during a printout
your pointer depth is 0
This is meant to simplify the way LLDB shows data on screen for small structs and similarly compact data types (e.g. std::pair<int,int> anyone?)
Feedback is especially welcome on how the feature feels and corner cases where we should apply this printout and don't (or viceversa, we are applying it when we shouldn't be)
llvm-svn: 191996
that all clients use them explicitly. This will hopefully
prevent any future confusion where things get cast to types
we don't expect.
<rdar://problem/15146458>
llvm-svn: 191984
to be explicit, to prevent horrid things like
std::string a = ConstString("foo")
from taking the path ConstString -> bool -> char
-> std::string.
This fixes, among other things, ClangFunction.
<rdar://problem/15137989>
llvm-svn: 191934
DumpValueObject() 2.0
This checkin restores pre-Xcode5 functionality to the "po" (expr -O) command:
- expr now has a new --description-verbosity (-v) argument, which takes either compact or full as a value (-v is the same as -vfull)
When the full mode is on, "po" will show the extended output with type name, persistent variable name and value, as in
(lldb) expr -O -v -- foo
(id) $0 = 0x000000010010baf0 {
1 = 2;
2 = 3;
}
When -v is omitted, or -vcompact is passed, the Xcode5-style output will be shown, as in
(lldb) expr -O -- foo
{
1 = 2;
2 = 3;
}
- for a non-ObjectiveC object, LLDB will still try to retrieve a summary and/or value to display
(lldb) po 5
5
-v also works in this mode
(lldb) expr -O -vfull -- 5
(int) $4 = 5
On top of that, this is a major refactoring of the ValueObject printing code. The functionality is now factored into a ValueObjectPrinter class for easier maintenance in the future
DumpValueObject() was turned into an instance method ValueObject::Dump() which simply calls through to the printer code, Dump_Impl has been removed
Test case to follow
llvm-svn: 191694
line breakpoints past the prologue of functions so it can be shared between the
file & line breakpoint resolver, and the source pattern breakpoint resolver,
and then share it.
llvm-svn: 191478
not have breakpoints set on it inserted into code that does have a valid line number. So allow
that line number, and the ThreadPlanStepRange should just continue stepping over 0 line ranges
as if they had the same line number as whatever we were previously stepping through.
llvm-svn: 191477
Now that SBValues can be setup to ignore synthetic values, this is no longer necessary, and so m_suppress_synthetic_value can go away
Another Hack Bites the Dust
llvm-svn: 191338
to build out the symbol table as addresses are used, and implements
the mechanism for ELF to add stripped symbols from eh_frame.
Uses this mechanism to allow disassembly for addresses corresponding
to stripped symbols for ELF, and provide hooks to implement this for
PE COFF.
Also removes eSymbolContextTailCall in favor of an option for
ResolveSymbolContextForAddress for consistency with the documentation
for eSymbolContextEverything. Essentially, this is just an option for
interpreting the so_addr.
llvm-svn: 191307
Specifically, allows the unwinder to handle the case where sc.function
gets resolved with a pc that is one past the address range of the function
(consistent with a tail call). However, there is no matching symbol.
Adds eSymbolContextTailCall to provide callers with control over the scope
of symbol resolution and to allow ResolveSymbolContextForAddress to handle
tail calls since this routine is common to unwind and disassembly.
llvm-svn: 191102
- searches frames beginning from the current frame, stops when an equivalent context is found
- not using GetStackFrameCount() for performance reasons
- fixes TestInlineStepping (clang/gcc buildbots)
llvm-svn: 190868
for the frame is one past the address range of the calling function.
- Lowers the fix from RegisterContextLLDB for use with disassembly
- Fixes one of three issues in the disassembly test in TestInferiorAssert.py
Also adds documentation that explains the resolution depths and interface.
Note: This change affects the resolution scope for eSymbolContextFunction
without impacting the performance of eSymbolContextSymbol.
Thanks to Matt Kopec for his review.
llvm-svn: 190812
with prefer_file_cache == false. This is what we want to do when
the user is doing a disassemble command -- show the actual memory
contents in case the memory has been corrupted or something -- but
when we're profiling functions for stepping or unwinding
(ThreadPlanStepRange::GetInstructionsForAddress,
UnwindAssemblyInstEmulation::GetNonCallSiteUnwindP) we can read
__TEXT instructions directly out of the file, if it exists.
<rdar://problem/14397491>
llvm-svn: 190638
This allows the PC to be directly changed to a different line.
It's similar to the example python script in examples/python/jump.py, except implemented as a builtin.
Also this version will track the current function correctly even if the target line resolves to multiple addresses. (e.g. debugging a templated function)
llvm-svn: 190572
SVN r189964 provided a sample Python script to inspect unordered(multi){set|map} with synthetic children, contribued by Jared Grubb
This checkin converts that sample script to a C++ provider built into LLDB
A test case is also provided
llvm-svn: 190564
setting of the environment variable COMMAND_MODE. Changed the Platform::GetResumeCountForShell
to Platform::GetResumeCountForLaunchInfo, and check both the shell and in the case of
/bin/sh the environment as well.
llvm-svn: 190538
that /bin/sh re-exec's itself to /bin/bash, so it needs one more resume when you
are using it as the shell than /bin/bash did or you will stop at the start of your
program, rather than running it.
So I added a Platform API to get the number of resumes needed when launching with
a particular shell, and set the right values for Mac OS X.
<rdar://problem/14935282>
llvm-svn: 190381
/bin/sh is more portable, and all systems with /bin/bash are expected to
have /bin/sh as well, even if only a link to bash.
Review: http://llvm-reviews.chandlerc.com/D1576
llvm-svn: 189879
- mode_t is defined in <sys/types.h>
- reorganized S_* user rights into win32.h
- Use Host::Kill instead of kill
- Currently #ifdef functions using pread/pwrite.
llvm-svn: 189364
Since I renamed most of the LLVM Mach-O enums in r189314, I had to go fix
LLDB to use the new names. While I was here, I decided that a COFF
plugin really shouldn't be using Mach-O enums.
llvm-svn: 189316
Summary:
This merge brings in the improved 'platform' command that knows how to
interface with remote machines; that is, query OS/kernel information, push
and pull files, run shell commands, etc... and implementation for the new
communication packets that back that interface, at least on Darwin based
operating systems via the POSIXPlatform class. Linux support is coming soon.
Verified the test suite runs cleanly on Linux (x86_64), build OK on Mac OS
X Mountain Lion.
Additional improvements (not in the source SVN branch 'lldb-platform-work'):
- cmake build scripts for lldb-platform
- cleanup test suite
- documentation stub for qPlatform_RunCommand
- use log class instead of printf() directly
- reverted work-in-progress-looking changes from test/types/TestAbstract.py that work towards running the test suite remotely.
- add new logging category 'platform'
Reviewers: Matt Kopec, Greg Clayton
Review: http://llvm-reviews.chandlerc.com/D1493
llvm-svn: 189295
address.
When loading a dSYM, and the file addresses of the dSYM Sections are
different than the executable binary Sections' file addresses, the
debug info won't be remapped to the actual load addresses correctly.
This only happens with binaries on the in-memory shared cache binaries
where their File addresses have been set to their actual load address
(outside an offset value) whereas the original executable and dSYM
have 0-based File addresses.
I think this patch will not be activated for other cases -- this is
the only case we know of where the dSYM and the executable's File
addresses differ -- but if this causes other problems we can restrict
it more carefully.
<rdar://problem/12335086>
llvm-svn: 188532
LLDB needs in memory module load level settings to control how much information is read from memory when loading in memory modules. This change adds a new setting:
(lldb) settings set target.memory-module-load-level [minimal|partial|complete]
minimal will load only sections (no symbols, or function bounds via function starts or EH frame)
partial will load sections + bounds
complete will load sections + bounds + symbols
llvm-svn: 188246
- Immediates can be shown as hex (either Intel or MASM style)
- See TestSettings.py for usage examples
- Verified to cause no regressions on Linux x86_64 (Ubuntu 12.10)
Patch by Richard Mitton!
llvm-svn: 187921
Fixed a crasher when using memory threads where a thread is sticking around too long and was causing problems when it didn't have a thread plan.
llvm-svn: 187395
LLDB requires that the inferior process be stopped before, and remain
stopped during, certain accesses to process state.
Previously this was achieved with a POSIX rwlock which had a write lock
taken for the duration that the process was running, and released when
the process was stopped. Any access to process state was performed with
a read lock held.
However, POSIX requires that pthread_rwlock_unlock() be called from the
same thread as pthread_rwlock_wrlock(), and lldb needs to stop and start
the process from different threads. Violating this constraint is
technically undefined behaviour, although as it happens Linux and Darwin
result in the unlock proceeding in this case. FreeBSD follows POSIX
more strictly, and the unlock would fail, resulting in a hang later upon
the next attempt to take the lock.
All read lock consumers use ReadTryLock() and handle failure to obtain
the lock (typically by logging an error "process is running"). Thus,
instead of using the lock state itself to track the running state, this
change adds an explicit m_running flag. ReadTryLock tests the flag, and
if the process is not running it returns with the read lock held.
WriteLock and WriteTryLock are renamed to SetRunning and TrySetRunning,
and (if successful) they set m_running with the lock held. This way,
read consumers can determine if the process is running and act
appropriately, and write consumers are still held off from starting the
process if read consumers are active.
Note that with this change there are still some curious access patterns,
such as calling WriteUnlock / SetStopped twice in a row, and there's no
protection from multiple threads trying to simultaneously start the
process. In practice this does not seem to be a problem, and was
exposing other undefined POSIX behaviour prior to this change.
llvm-svn: 187377
Greg Clayton