Allowing LLDB to resolve names of Python functions when they are located in classes
This allows things like *bound* classmethods to be used for formatters, commands, ...
llvm-svn: 183772
Adding a new setting interpreter.stop-command-source-on-error that dictates a default behavior for whether command source should stop upon hitting an error
You can still override the setting for each individual invocation with the usual -e setting
llvm-svn: 183719
Add support for half-floats, as specified by IEEE-754-2008
With this checkin, you can now say:
(lldb) x/7hf foo
to read 7 half-floats at address foo
llvm-svn: 183716
Reworked the download information on lldb.llvm.org:
- svn copy of download.html to source.html with information on source-code access
- new download.html with links to download nightly builds and Debian releases
- updated the sidebar to reflect these changes
llvm-svn: 183547
level. Fixes a bug in "break set --source-pattern-regexp" when a shared library is
specified.
Also cleaned up the help text for --source-pattern-regexp so it is a little clearer.
<rdar://problem/14084261>
llvm-svn: 183476
lldb doesn't autocomplete objective C class methods. The regular expression was looking for strings that started with the completion string that was passed in. For objective C class methods, this string starts with "+" which wasn't being escaped. Added many other escapes that were missing just in case.
llvm-svn: 183470
you can now specify:
debugserver host:port
debugserver port
debugserver /path/to/file
When "host" is specified, we will only accept connections from that host. If host is not specified, we default to "localhost".
llvm-svn: 183457
condition in two different processes (with the
same target) could cause crashes. Now the breakpoint
condition is always evaluated (and possibly parsed)
by one thread at a time.
<rdar://problem/14083737>
llvm-svn: 183440
that is patterned after its parent TestInferiorCrashing.py.
- The xfail decorator limits the xfail to tool-chains that support this compiler option.
- Included a TODO concerning the platform-specific behavior when 'next' is issued after a crash.
- Toggling -fomit-frame-pointer results in an xpass as mentioned in pr15415.
Thanks to Daniel for the review, and Samuel for the bug report and reproducer.
llvm-svn: 183434
- For instance, allows 'gcc' to match x86-64-linux-gnu-gcc as required on some Debian builds.
- Also adds doc-strings and a more consistent naming convention for related helpers.
llvm-svn: 183415
- Implemented the SExt instruction, and
- eliminated redundant codepaths for constant
handling.
Added test cases.
<rdar://problem/13244258>
<rdar://problem/13955820>
llvm-svn: 183344
- one test case is due to llvm.org/pr16229
- other test case uses a Linux workaround for above by using os.fork() instead of subprocess module
Patch by Andy Kaylor!
llvm-svn: 183340
Adding data formatters for std::set, std::multiset and std::multimap for libc++
The underlying data structure is the same as std::map, so this change is very minimal and mostly consists of test cases
llvm-svn: 183323
- Ensures that this container is populated once for the lifetime of lldb
--- In particular, static methods can query this data even after the first RegisterContext has been destroyed.
- Uses a singleton function to avoid global constructors.
Thanks to Greg Clayton for the suggestion!
llvm-svn: 183313
the link register save location being in the link register - in which case we
should iterate down the stack, not recursively try to find the lr in the current
frame over and over.
<rdar://problem/13932954>
llvm-svn: 183282
Two things:
1) fixing a bug where memory read was not clearing the m_force flag after it was passed, so that subsequent memory reads would not need to be forced even if over boundary
2) adding a setting target.max-memory-read-size that you can set instead of the hardcoded 1024 bytes limit we had before
llvm-svn: 183276
If you want to define a formatter for "array of Foo of any size", ordinarily you would say
-x "Foo \[[0-9]+\]"
this checkin allows you to instead say "Foo[]" (or "Foo []") and LLDB will automatically create the regular expression and add the -x flag on your behalf
llvm-svn: 183272
Accept mach-o files with bad segments. Many core files are not created correctly and we should still be able to glean any information we can from them.
llvm-svn: 183247
LLDB API versioning
This checkin makes the LLDB API versioned
We are starting at version 1.0 and will then revise and update the API from there
Further details:
API versioning
---------------------------------
The LLDB API is versioned independently of the LLDB source base
Our API version numbers are composed of a major and a minor number
The major number means a complete and stable revision of the API. Major numbers are compatibility breakers
(i.e. when we change the API major number, there is no promise of compatibility with the previous major version
and we are free to remove and/or change any APIs)
Minor numbers are a work-in-progress evolution of the API. APIs will not be removed or changed across minor versions
(minors do not break compatibility). However, we can deprecate APIs in minor versions or add new APIs in minor versions
A deprecated API is supposedly going to be removed in the next major version and will generate a warning if used
APIs we add in minor versions will not be removed (at least until the following major) but they might theoretically be deprecated
in a following minor version
Users are discouraged from using the LLDB version number to test for API features and should instead use the API version checking
as discussed below
API version checking
---------------------------------
You can (optionally) sign into an API version checking feature
To do so you need to define three macros:
LLDB_API_CHECK_VERSIONING - define to any value (or no value)
LLDB_API_MAJOR_VERSION_WANTED - which major version of the LLDB API you are targeting
LLDB_API_MINOR_VERSION_WANTED - which minor version of the LLDB API you are targeting
If these macros exist - LLDB will enable version checking of the public API
If LLDB_API_MAJOR_VERSION is not equal to LLDB_API_MAJOR_VERSION_WANTED we will immediately halt your compilation with an error
This is by design, since we do not make any promise of compatibility across major versions - if you really want to test your luck, disable the versioning altogether
If the major version test passes, you have signed up for a specific minor version of the API
Whenever we add or deprecate an API in a minor version, we will mark it with either
LLDB_API_NEW_IN_DOT_x - this API is new in LLDB .x
LLDB_API_DEPRECATED_IN_DOT_x - this API is deprecated as of .x
If you are using an API new in DOT_x
if LLDB_API_MINOR_VERSION_WANTED >= x then all is well, else you will get a compilation error
This is meant to prevent you from using APIs that are newer than whatever LLDB you want to target
If you are using an API deprecated in DOT_x
if LLDB_API_MINOR_VERSION_WANTED >= x then you will get a compilation warning, else all is well
This is meant to let you know that you are using an API that is deprecated and might go away
Caveats
---------------------------------
Version checking only works on clang on OSX - you will get an error if you try to enable it on any other OS/compiler
If you want to enable version checking on other platforms, you will need to define appropriate implementations for
LLDB_API_IMPL_DEPRECATED and LLDB_API_IMPL_TOONEW and any other infrastructure your compiler needs for this purpose
We have no deprecation-as-error mode
There is no support for API versioning in Python
We reserve to use macros whose names begin with LLDB_API_ and you should not use them in your source code as they might conflict
with present or future macro names we are using to implement versioning
For API implementors:
If you need to add a new public API call, please remember to add the LLDB_API_NEW_IN_DOT_x marker in the header file
and when you are done with adding stuff, to also update LLDB_API_MINOR_VERSION
If you want to remove a function, deprecate it first, by using LLDB_API_DEPRECATED_IN_DOT_x
and when you are done with deprecating stuff, to also update LLDB_API_MINOR_VERSION
A new major version (LLDB_API_MAJOR_VERSION++) is your only chance to remove and/or change API calls
but is probably quite a big deal and you might want to consider deprecating the existing calls for a while
before doing your changes
A couple more caveats:
Currently, the lldb-tool does NOT use the version checking feature. It would be a nice future improvement to make it do that, once we have proper version checking on other OSs
APIs marked as deprecated by a comment in the source are still deprecated just that way. A good purpose for API 1.1 might be to deprecate them with appropriate markers
llvm-svn: 183244
Fixing an issue where formats would not propagate from parents to children in all cases
Details follow:
an SBValue has children and those are fetched along with their values
Now, one calls SBValue::SetFormat() on the parent
Technically, the format choices should propagate onto the children (see ValueObject::GetFormat())
But if the children values are already fetched, they won't notice the format change and won't update themselves
This commit fixes that by making ValueObject::GetValueAsCString() check if any format change intervened from the previous call to the current one
A test case is also added
llvm-svn: 183030
command script import now does reloads - for real
If you invoke command script import foo and it detects that foo has already been imported, it will
- invoke reload(foo) to reload the module in Python
- re-invoke foo.__lldb_init_module
This second step is necessary to ensure that LLDB does not keep cached copies of any formatter, command, ... that the module is providing
Usual caveats with Python imports persist. Among these:
- if you have objects lurking around, reloading the module won't magically update them to reflect changes
- if module A imports module B, reloading A won't reload B
These are Python-specific issues independent of LLDB that would require more extensive design work
The --allow-reload (-r) option is maintained for compatibility with existing scripts, but is clearly documented as redundant - reloading is always enabled whether you use it or not
llvm-svn: 182977
Upon encountering an object not of type string, LLDB will get the string representation of it (akin to calling str(X) in Python code) and use that as the summary to display
Feedback is welcome as to whether repr() should be used instead (but the argument for repr() better be highly persuasive :-)
llvm-svn: 182953
Cleaned up the thread updating code in the OperatingSystemPython class. It doesn't need to clear the "new_thread_list" anymore as it is always empty.
It also now assigns the "core_thread_list" to "new_thread_list" if no threads are detected through python.
llvm-svn: 182893
- The original test now passes on Linux with clang because a breakpoint is hit prior to evaluation of text_list, which improves text coverage.
- The new test fails because 4 steps are requested, and only two occur prior to evaluation of text_list.
--- Note that the loss of every second "next" command can be reproduced using lldb manually with this script.
llvm-svn: 182860
- added code for tracking transition from eStateAttaching to eStateStopped in event listener and handling process continuation there.
Patch by Arthur Evstifeev!
llvm-svn: 182806
Giving a timeout for the call to NSPrintForDebugger() that happens when you “po” objects
This is a temporary workaround until a more detailed solution to the general problem of canceling actions is found
llvm-svn: 182782
- [ninja|make] lldb-cpp-doc builds the C++ API reference docs
- [ninja|make] lldb-python-doc builds the python API reference docs
- updated build page on website to include instructions to build docs
Tested on Linux/Mac OS X
llvm-svn: 182752
- Fix for attach by name
- Details for register support
- Punted on i386 details as its status has drifted since this page was originally posted
- Multi-threaded target support is soon to be released on Linux
- Partial back-trace is called out since its a high-profile issue
llvm-svn: 182664
Fixed performance issues that arose after changing SBTarget, SBProcess, SBThread and SBFrame over to using a std::shared_ptr to a ExecutionContextRef. The ExecutionContextRef doesn't store a std::weak_ptr to a stack frame because stack frames often get replaced with new version, so it held onto a StackID object that would allow us to ask the thread each time for the frame for the StackID. The linear function was too slow for large recursive stacks. We also fixed an issue where anytime the std::shared_ptr<ExecutionContextRef> in any SBTarget, SBProcess, SBThread objects was turned into an ExecutionContext object, it would try to resolve all items in the ExecutionContext which are shared pointers. Even if the StackID in the ExecutionContextRef was invalid, it was looking through all frames in every thread. This causes a lot of unnecessary frame accesses.
llvm-svn: 182627
1. Added new :Lpo command
2. :Lpo and :Lprint can be invoked without parameters. In that case
cursor word will be used
3. Added :LpO command in that case instead of <cword> will be used
stripped <cWORD>. This command is useful for printing objective-c
properties (for ex.: self.tableView).
Patch by Arthur Evstifeev!!
llvm-svn: 182613
Which means "platform process list" should work and list the architecture.
We are now parsing the elf build-id if it exists, which should allow us to load stripped symbols (looking at that next).
llvm-svn: 182610
settings set use-color [false|true]
settings set prompt "${ansi.bold}${ansi.fg.green}(lldb)${ansi.normal} "
also "--no-use-colors" on the command prompt
llvm-svn: 182609
removed the bitfields. This should be conforming
C++11, though, cf. C++03 9.6(3):
"
A bit-field shall have integral or enumeration
type (3.9.1).
"
llvm-svn: 182545
Added logging for the OS plug-in python objects in OperatingSystemPython so we can see the python dictionary returned from the plug-in when logging is enabled.
llvm-svn: 182530
live as long as they needed to. This led to
equality tests involving persistent variables
often failing or succeeding when they had no
business doing so.
To do this, I introduced the ability for a
memory allocation to "leak" - that is, to
persist in the process beyond the lifetime of
the expression. Hand-declared persistent
variables do this now.
<rdar://problem/13956311>
llvm-svn: 182528
Fixed ProcessMachCore to be able to locate the main executeable in the core file even if it doesn't start at a core file address range boundary. Prior to this we only checked the first bytes of each range in the core file for mach_kernel or dyld. Now we still do this, but if we don't find the mach_kernel or dyld anywhere, we go through all core file ranges and check every 0x1000 to see if we can find dyld or the mach_kernel.
Now that we can properly detect the mach_kernel at any address, we don't need to call "DynamicLoaderDarwinKernel::SearchForDarwinKernel(Process*)" anymore.
llvm-svn: 182513
Lock the lldb_private::Module mutex while tearing down the module to make sure we don't get clients accessing the contents on a module as it is going away.
llvm-svn: 182511
Another fix to make sure that if we aren't able to extract an object file for any reason, we don't crash when trying to parse the debug map info.
llvm-svn: 182441
Yet another implementation of the python in dSYM autoload :)
This time we are going with a ternary setting:
true - load, do not warn
false - do not load, do not warn
warn - do not load, warn (default)
llvm-svn: 182414
A user request such as: memory read -fc -s10 -c1 *charPtrPtr would cause us to crash upon trying to read 1 char of size 10 from memory
This request is now translated into: memory read -fc -s1 -c10 *charPtrPtr (i.e. read 10 chars of size 1 from memory) which is probably also what the user originally wanted
llvm-svn: 182398
There are two settings:
target.load-script-from-symbol-file is a boolean that says load or no load (default: false)
target.warn-on-script-from-symbol-file is also a boolean, it says whether you want to be warned when a script file is not loaded due to security (default: true)
the auto loading on change for target.load-script-from-symbol-file is preserved
llvm-svn: 182336
This changes the setting target.load-script-from-symbol-file to be a ternary enum value:
default (the default value) will NOT load the script files but will issue a warning suggesting workarounds
yes will load the script files
no will not load the script files AND will NOT issue any warning
if you change the setting value from default to yes, that will then cause the script files to be loaded
(the assumption is you didn't know about the setting, got a warning, and quickly want to remedy it)
if you have a settings set command for this in your lldbinit file, be sure to change "true" or "false" into an appropriate "yes" or "no" value
llvm-svn: 182323
Name matching was working inconsistently across many places in LLDB. Anyone doing name lookups where you want to look for all types of names should used "eFunctionNameTypeAuto" as the sole name type mask. This will ensure that we get consistent "lookup function by name" results. We had many function calls using as mask like "eFunctionNameTypeBase | eFunctionNameTypeFull | eFunctionNameTypeMethod | eFunctionNameTypeSelector". This was due to the function lookup by name evolving over time, but as it stands today, use eFunctionNameTypeAuto when you want general name lookups. Either ModuleList::FindFunctions() or Module::FindFunctions() will figure out the right kinds of names to lookup and remove the "eFunctionNameTypeAuto" and replace it with the exact subset of what the name can be.
This checkin also changes eFunctionNameTypeAny over to use eFunctionNameTypeAuto to reflect this.
llvm-svn: 182179
- copy lldb python module into directory specified with CMAKE_INSTALL_PREFIX
- make liblldb.so a symlink (to liblldb.so.X.Y where X.Y is the LLVM version)
llvm-svn: 182157
- Note that this is not correct, as the failure is associated with build options of libc.so, however it's failing on a Debian buildbot that uses gcc 4.6.2 (and the real goal is a complete backtrace even with -fomit-frame-pointer).
- Adds helpers to lldbtest.py to check the expectedCompiler and expectedVersion, with an eventual goal of reducing the number of test decorators.
--- Currently allows a comparison operator and a compiler version to be specified.
--- Can be extended to support ranges of compiler versions.
llvm-svn: 182155
-Remove tracing of fork/vfork until we add support for tracing inferiors' children on Linux.
-Add trace exec option for ptrace so that we don't receive legacy SIGTRAP signals on execve calls.
-Add handling of SIGCHLD sent by kernel (for now, deliver the signal to the inferior).
llvm-svn: 182153
- now, the output binary is called 'lldb-3.4' instead of 'lldb'
- a symlink 'lldb' -> 'lldb-3.4' is also created
- this fixes one of the problems preventing CMake from building Debian packages
llvm-svn: 182148
- On Linux, the partial back-trace after an assert can cause the basic test to fail as discussed on lldb-dev.
- Uses SBFrame to walk up the stack to the assert site and tests expression evaluation of locals, globals and arguments.
Thanks to Daniel for review and testing on OS/X.
llvm-svn: 182115
"source list -n <func>" can now show more than one location that matches a function name. It will unique multiple of the same source locations so they don't get displayed. It also handles inline functions correctly.
llvm-svn: 182067
Show variables that were in the debug info but optimized out. Also display a good error message when one of these variables get used in an expression.
llvm-svn: 182066
regions that aren't actually allocated in the
process. This cache is used by the expression
parser if the underlying process doesn't support
memory allocation, to avoid needless repeated
searches for unused address ranges.
Also fixed a silly bug in IRMemoryMap where it
would continue searching even after it found a
valid region.
<rdar://problem/13866629>
llvm-svn: 182028
Make type summary add and breakpoint command add show an helpful prototype + argument reference when manually typing Python code for these elements
llvm-svn: 181968
Fixed "target symbols add" to correctly extract all module specifications from a dSYM file that is supplied and match the symbol file to a current target module using the UUID values if they are available.
This fixes the case where you add a dSYM file (like "foo.dSYM") which is for a renamed executable (like "bar"). In our case it was "mach_kernel.dSYM" which didn't match "mach_kernel.sys".
llvm-svn: 181916
- s/skipOnLinux/skipIfLinux/ to match style of every other decorator
- linkify bugizilla/PR numbers in comments
No intended change in functionality.
llvm-svn: 181913
- newlines from GetRepositoryPath output were interfering with ninja builds
- replace newlines with spaces
- remove *only* trailing spaces from repo path
llvm-svn: 181899
Python breakpoint actions can return False to say that they don't want to stop at the breakpoint to which they are associated
Almost all of the work to support this notion of a breakpoint callback was in place, but two small moving parts were missing:
a) the SWIG wrapper was not checking the return value of the script
b) when passing a Python function by name, the call statement was dropping the return value of the function
This checkin addresses both concerns and makes this work
Care has been taken that you only keep running when an actual value of False has been returned, and that any other value (None included) means Stop!
llvm-svn: 181866
process StopLocker (if there is a process) before it will hand out SBValues. We were doing this in
an ad hoc fashion previously, and then playing whack-a-mole whenever we found a place where we should
have been doing this but weren't. Really, it doesn't make sense to be poking at SBValues when the target
is running, the dynamic and synthetic values can't really be computed, and the underlying memory may be
incoherent.
<rdar://problem/13819378> Sometimes when stepping fast, my inferior is killed by debugserver
llvm-svn: 181863
Combine N_GSYM stab entries with their non-stab counterpart (data symbols) to make the symbol table smaller with less duplicate named symbols.
llvm-svn: 181841
- test_breakpoint_callback -- filed llvm.org/pr-16000
- test_listener_resume -- resume a process from a thread waiting on SBListener
- test_listener_event_description -- SBEvent description from SBListener thread
- test_listener_event_process -- query process/thread/stack info from SBListener thread
llvm-svn: 181819
The "lldb" driver was interfering with STDOUT and STDERR if the output was over 1024 charcters long. The output was grabbing 1024 characters at a time, before it output the characters, it was writing characters to the screen to clear the current line. This has been fixed.
I also fixed the command interpreter from mixing the "(lldb) " prompt in with program output by always manually checking for program output. This was done by having the command interpreter know when it is in the middle of executing a command by setting a bool. This was needed since sometimes when a command would run the target, like with a command like 'expression (int)printf("hello\n")', the process would push a new input reader, and then pop it when it was done. This popping of the input reader would cause the command interpreter to get sent a reactivated message (from the private process state thread) and cause it to ask for another command, even though we were still in the middle of the command ('expression (int)printf("hello\n")'). Now we set a bool to true, run the command and set the bool to false. If we get reactivated while we are in the middle of a command, we don't say we are ready for a new command. This coupled with emitting the STDOUT/STDERR first after each command, followed by the command results, followed by then saying we are ready for a new command, should help cleanup the command line output on all platforms.
llvm-svn: 181807
- add IsVirtualStep() virtual function to ThreadPlan, and implement it for
ThreadPlanStepInRange
- make GetPrivateStopReason query the current thread plan for a virtual stop to
decide if the current stop reason needs to be preserved
- remove extra check for an existing process in GetPrivateStopReason
llvm-svn: 181795
Most importantly, have DoReadGPR/DoReadFPU/DoReadEXC return -1
to indicate failure if they're called. Else these could override
the Error setting for the relevant thread state -- if the core file
didn't include a floating point thread state, for instance, these
functions would clear the Error setting for that register set and
lldb would display random bytes as those registers' contents.
<rdar://problem/13665075>
llvm-svn: 181757
Enrico Granata