DWARF with .o files now uses 40-60% less memory!
Big fixes include:
- Change line table internal representation to contain "file addresses". Since each line table is owned by a compile unit that is owned by a module, it makes address translation into lldb_private::Address easy to do when needed.
- Removed linked address members/methods from lldb_private::Section and lldb_private::Address
- lldb_private::LineTable can now relink itself using a FileRangeMap to make it easier to re-link line tables in the future
- Added ObjectFile::ClearSymtab() so that we can get rid of the object file symbol tables after we parse them once since they are not needed and kept memory allocated for no reason
- Moved the m_sections_ap (std::auto_ptr to section list) and m_symtab_ap (std::auto_ptr to the lldb_private::Symtab) out of each of the ObjectFile subclasses and put it into lldb_private::ObjectFile.
- Changed how the debug map is parsed and stored to be able to:
- Lazily parse the debug map for each object file
- not require the address map for a .o file until debug information is linked for a .o file
llvm-svn: 176454
lldb was mmap'ing archive files once per .o file it loads, now it correctly shares the archive between modules.
LLDB was also always mapping entire contents of universal mach-o files, now it maps just the slice that is required.
Added a new logging channel for "lldb" called "mmap" to help track future regressions.
Modified the ObjectFile and ObjectContainer plugin interfaces to take a data offset along with the file offset and size so we can implement the correct caching and efficient reading of parts of files without mmap'ing the entire file like we used to.
The current implementation still keeps entire .a files mmaped (once) and entire slices from universal files mmaped to ensure that if a client builds their binaries during a debug session we don't lose our data and get corrupt object file info and debug info.
llvm-svn: 174524
The first part of the fix for having LLDB handle LTO debugging when the DWARF is in the .o files. This part separates the object file's modules into a separate cache map that maps unique C strings for the N_OSO path to the ModuleSP since one object file might be mentioned more than once in LTO binaries.
llvm-svn: 174476
Fix in loading mach files from memory when using DynamicLoaderMacOSXDYLD.
Removed the uuid mismatch warning that could be spit out and any time during debugging and removed the test case that was looking for that. Currently the "add-dsym" or "target symbols add" command will report an error when the UUID's don't match.
Be more careful when checking and resolving section + offset addresses to make sure none of the base addresses are invalid.
llvm-svn: 174222
Simplify the logging on ObjectFile::~ObjectFile() to not access an classes above the object file (like the module) so we don't crash when logging object lifetimes. The log message contains the "this" pointer value which can be matched up with the constructor log.
llvm-svn: 168754
it is unconditionally present now.
ObjectContainerBSDArchive::CreateInstance %z8.8x is not a valid printf arg specifier, %8.8zx would work
for size_t arg but this arg is addr_t. use %8.8llx and cast up to uint64_t.
ObjectFile::FindPlugin ditto.
DynamicRegisterInfo::SetRegisterInfo ifdef this function out if LLDB_DISABLE_PYTHON.
llvm-svn: 163599
Make breakpoint setting by file and line much more efficient by only looking for inlined breakpoint locations if we are setting a breakpoint in anything but a source implementation file. Implementing this complex for a many reasons. Turns out that parsing compile units lazily had some issues with respect to how we need to do things with DWARF in .o files. So the fixes in the checkin for this makes these changes:
- Add a new setting called "target.inline-breakpoint-strategy" which can be set to "never", "always", or "headers". "never" will never try and set any inlined breakpoints (fastest). "always" always looks for inlined breakpoint locations (slowest, but most accurate). "headers", which is the default setting, will only look for inlined breakpoint locations if the breakpoint is set in what are consudered to be header files, which is realy defined as "not in an implementation source file".
- modify the breakpoint setting by file and line to check the current "target.inline-breakpoint-strategy" setting and act accordingly
- Modify compile units to be able to get their language and other info lazily. This allows us to create compile units from the debug map and not have to fill all of the details in, and then lazily discover this information as we go on debuggging. This is needed to avoid parsing all .o files when setting breakpoints in implementation only files (no inlines). Otherwise we would need to parse the .o file, the object file (mach-o in our case) and the symbol file (DWARF in the object file) just to see what the compile unit was.
- modify the "SymbolFileDWARFDebugMap" to subclass lldb_private::Module so that the virtual "GetObjectFile()" and "GetSymbolVendor()" functions can be intercepted when the .o file contenst are later lazilly needed. Prior to this fix, when we first instantiated the "SymbolFileDWARFDebugMap" class, we would also make modules, object files and symbol files for every .o file in the debug map because we needed to fix up the sections in the .o files with information that is in the executable debug map. Now we lazily do this in the DebugMapModule::GetObjectFile()
Cleaned up header includes a bit as well.
llvm-svn: 162860
1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags
2 - symbol names are marked as mangled if they start with "_Z"
Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF.
llvm-svn: 153496
This fix really needed to happen as a previous fix I had submitted for
calculating symbol sizes made many symbols appear to have zero size since
the function that was calculating the symbol size was calling another function
that would cause the calculation to happen again. This resulted in some symbols
having zero size when they shouldn't. This could then cause infinite stack
traces and many other side affects.
llvm-svn: 152244
I started work on being able to add symbol files after a debug session
had started with a new "target symfile add" command and quickly ran into
problems with stale Address objects in breakpoint locations that had
lldb_private::Section pointers into modules that had been removed or
replaced. This also let to grabbing stale modules from those sections.
So I needed to thread harded the Address, Section and related objects.
To do this I modified the ModuleChild class to now require a ModuleSP
on initialization so that a weak reference can created. I also changed
all places that were handing out "Section *" to have them hand out SectionSP.
All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild
so all of the find plug-in, static creation function and constructors now
require ModuleSP references instead of Module *.
Address objects now have weak references to their sections which can
safely go stale when a module gets destructed.
This checkin doesn't complete the "target symfile add" command, but it
does get us a lot clioser to being able to do such things without a high
risk of crashing or memory corruption.
llvm-svn: 151336
user space programs. The core file support is implemented by making a process
plug-in that will dress up the threads and stack frames by using the core file
memory.
Added many default implementations for the lldb_private::Process functions so
that plug-ins like the ProcessMachCore don't need to override many many
functions only to have to return an error.
Added new virtual functions to the ObjectFile class for extracting the frozen
thread states that might be stored in object files. The default implementations
return no thread information, but any platforms that support core files that
contain frozen thread states (like mach-o) can make a module using the core
file and then extract the information. The object files can enumerate the
threads and also provide the register state for each thread. Since each object
file knows how the thread registers are stored, they are responsible for
creating a suitable register context that can be used by the core file threads.
Changed the process CreateInstace callbacks to return a shared pointer and
to also take an "const FileSpec *core_file" parameter to allow for core file
support. This will also allow for lldb_private::Process subclasses to be made
that could load crash logs. This should be possible on darwin where the crash
logs contain all of the stack frames for all of the threads, yet the crash
logs only contain the registers for the crashed thrad. It should also allow
some variables to be viewed for the thread that crashed.
llvm-svn: 150154
Fixed "target modules list" (aliased to "image list") to output more information
by default. Modified the "target modules list" to have a few new options:
"--header" or "-h" => show the image header address
"--offset" or "-o" => show the image header address offset from the address in the file (the slide applied to the shared library)
Removed the "--symfile-basename" or "-S" option, and repurposed it to
"--symfile-unique" "-S" which will show the symbol file if it differs from
the executable file.
ObjectFile's can now be loaded from memory for cases where we don't have the
files cached locally in an SDK or net mounted root. ObjectFileMachO can now
read mach files from memory.
Moved the section data reading code into the ObjectFile so that the object
file can get the section data from Process memory if the file is only in
memory.
lldb_private::Module can now load its object file in a target with a rigid
slide (very common operation for most dynamic linkers) by using:
bool
Module::SetLoadAddress (Target &target, lldb::addr_t offset, bool &changed)
lldb::SBModule() now has a new constructor in the public interface:
SBModule::SBModule (lldb::SBProcess &process, lldb::addr_t header_addr);
This will find an appropriate ObjectFile plug-in to load an image from memory
where the object file header is at "header_addr".
llvm-svn: 149804
due to RTTI worries since llvm and clang don't use RTTI, but I was able to
switch back with no issues as far as I can tell. Once the RTTI issue wasn't
an issue, we were looking for a way to properly track weak pointers to objects
to solve some of the threading issues we have been running into which naturally
led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared
pointer from just a pointer, which is also easily solved using the
std::tr1::enable_shared_from_this class.
The main reason for this move back is so we can start properly having weak
references to objects. Currently a lldb_private::Thread class has a refrence
to its parent lldb_private::Process. This doesn't work well when we now hand
out a SBThread object that contains a shared pointer to a lldb_private::Thread
as this SBThread can be held onto by external clients and if they end up
using one of these objects we can easily crash.
So the next task is to start adopting std::tr1::weak_ptr where ever it makes
sense which we can do with lldb_private::Debugger, lldb_private::Target,
lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and
many more objects now that they are no longer using intrusive ref counted
pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive
pointers).
llvm-svn: 149207
mmap() the entire object file contents into memory with MAP_PRIVATE.
We do this because object file contents can change on us and currently
this helps alleviate this situation. It also make the code for accessing
object file data much easier to manage and we don't end up opening the
file, reading some data and closing the file over and over.
llvm-svn: 148017
object file can correctly make these symbols which will abstract us from the
file format and ABI and we can then ask for the objective C class symbol for
a class and find out which object file it was defined in.
llvm-svn: 145744
This is the actual fix for the above radar where global variables that weren't
initialized were not being shown correctly when leaving the DWARF in the .o
files. Global variables that aren't intialized have symbols in the .o files
that specify they are undefined and external to the .o file, yet document the
size of the variable. This allows the compiler to emit a single copy, but makes
it harder for our DWARF in .o files with the executable having a debug map
because the symbol for the global in the .o file doesn't exist in a section
that we can assign a fixed up linked address to, and also the DWARF contains
an invalid address in the "DW_OP_addr" location (always zero). This means that
the DWARF is incorrect and actually maps all such global varaibles to the
first file address in the .o file which is usually the first function. So we
can fix this in either of two ways: make a new fake section in the .o file
so that we have a file address in the .o file that we can relink, or fix the
the variable as it is created in the .o file DWARF parser and actually give it
the file address from the executable. Each variable contains a
SymbolContextScope, or a single pointer that helps us to recreate where the
variables came from (which module, file, function, etc). This context helps
us to resolve any file addresses that might be in the location description of
the variable by pointing us to which file the file address comes from, so we
can just replace the SymbolContextScope and also fix up the location, which we
would have had to do for the other case as well, and update the file address.
Now globals display correctly.
The above changes made it possible to determine if a variable is a global
or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag
for each variable (local, global, or static), yet DWARF provides no way for
us to classify these variables into these categories. We can now detect when
a variable has a simple address expressions as its location and this will help
us classify these correctly.
While making the above changes I also noticed that we had two symbol types:
eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same
thing: the symbol is not defined in the current object file. Symbol objects
also have a bit that specifies if a symbol is externally visible, so I got
rid of the eSymbolTypeExtern symbol type and moved all code locations that
used it to use the eSymbolTypeUndefined type.
llvm-svn: 144489
in the same hashed format as the ".apple_names", but they map objective C
class names to all of the methods and class functions. We need to do this
because in the DWARF the methods for Objective C are never contained in the
class definition, they are scattered about at the translation unit level and
they don't even have attributes that say the are contained within the class
itself.
Added 3 new formats which can be used to display data:
eFormatAddressInfo
eFormatHexFloat
eFormatInstruction
eFormatAddressInfo describes an address such as function+offset and file+line,
or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants).
The format character for this is "A", the long format is "address".
eFormatHexFloat will print out the hex float format that compilers tend to use.
The format character for this is "X", the long format is "hex float".
eFormatInstruction will print out disassembly with bytes and it will use the
current target's architecture. The format character for this is "i" (which
used to be being used for the integer format, but the integer format also has
"d", so we gave the "i" format to disassembly), the long format is
"instruction".
Mate the lldb::FormatterChoiceCriterion enumeration private as it should have
been from the start. It is very specialized and doesn't belong in the public
API.
llvm-svn: 143114
index the DWARF. Also fixed an issue with memory accelerator tables with a
size of 1 where we would loop infinitely.
Added support for parsing the new .apple_namespaces section which gives us a
memory hash table for looking up namespaces.
llvm-svn: 141128
used to do this because we needed to find the shared pointer for a .o
file when the .o file's module was needed in a SymbolContext since the
module in a symbol context was a shared pointer. Now that we are using
intrusive pointers we don't have this limitation anymore since any
instrusive shared pointer can be made from a pointer to an object
all on its own.
Also switched over to having the Module and SymbolVendor use shared
pointers to their object files as had a leak on MacOSX when the
SymbolVendor's object file wasn't the same as the Module's (debug info
in a stand along file (dSYM file)). Now everything will correctly clean
itself up when the module goes away after an executable gets rebuilt.
Now we correctly get rid of .o files that are used with the DWARF with
debug map executables on subsequent runs since the only shared pointer
to the object files in from the DWARF symbol file debug map parser, and
when the module gets replaced, it destroys to old one along with all .o
files.
Also added a small optimization when using BSD archives where we will
remove old BSD containers from the shared list when they are outdated.
llvm-svn: 140002
DWARF accelerator table sections to the DWARF parser. These sections are similar
to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables
that are saved to disc in a way that the sections can just be loaded into memory
and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes
sections are not ordered, contain a copy of the name in the section itself which
makes these sections quite large, they only include publicly exported names (so no
static functions, no types defined inside functions), many compilers put different
information in them making them very unreliable so most debugger ignore these sections
and parse the DWARF on their own. The tables must also be parsed and sorted in order
to be used effectively. The new sections can be quickly loaded and very efficiently be used
to do name to DIE lookups with very little up front work. The format of these new
sections will be changing while we work out the bugs, but we hope to have really
fast name to DIE lookups soon.
llvm-svn: 138979
the CommandInterpreter where it was always being used.
Make sure that Modules can track their object file offsets correctly to
allow opening of sub object files (like the "__commpage" on darwin).
Modified the Platforms to be able to launch processes. The first part of this
move is the platform soon will become the entity that launches your program
and when it does, it uses a new ProcessLaunchInfo class which encapsulates
all process launching settings. This simplifies the internal APIs needed for
launching. I want to slowly phase out process launching from the process
classes, so for now we can still launch just as we used to, but eventually
the platform is the object that should do the launching.
Modified the Host::LaunchProcess in the MacOSX Host.mm to correctly be able
to launch processes with all of the new eLaunchFlag settings. Modified any
code that was manually launching processes to use the Host::LaunchProcess
functions.
Fixed an issue where lldb_private::Args had implicitly defined copy
constructors that could do the wrong thing. This has now been fixed by adding
an appropriate copy constructor and assignment operator.
Make sure we don't add empty ModuleSP entries to a module list.
Fixed the commpage module creation on MacOSX, but we still need to train
the MacOSX dynamic loader to not get rid of it when it doesn't have an entry
in the all image infos.
Abstracted many more calls from in ProcessGDBRemote down into the
GDBRemoteCommunicationClient subclass to make the classes cleaner and more
efficient.
Fixed the default iOS ARM register context to be correct and also added support
for targets that don't support the qThreadStopInfo packet by selecting the
current thread (only if needed) and then sending a stop reply packet.
Debugserver can now start up with a --unix-socket (-u for short) and can
then bind to port zero and send the port it bound to to a listening process
on the other end. This allows the GDB remote platform to spawn new GDB server
instances (debugserver) to allow platform debugging.
llvm-svn: 129351
public types and public enums. This was done to keep the SWIG stuff from
parsing all sorts of enums and types that weren't needed, and allows us to
abstract our API better.
llvm-svn: 128239
platform status -- gets status information for the selected platform
platform create <platform-name> -- creates a new instance of a remote platform
platform list -- list all available platforms
platform select -- select a platform instance as the current platform (not working yet)
When using "platform create" it will create a remote platform and make it the
selected platform. For instances for iPhone OS debugging on Mac OS X one can
do:
(lldb) platform create remote-ios --sdk-version=4.0
Remote platform: iOS platform
SDK version: 4.0
SDK path: "/Developer/Platforms/iPhoneOS.platform/DeviceSupport/4.0"
Not connected to a remote device.
(lldb) file ~/Documents/a.out
Current executable set to '~/Documents/a.out' (armv6).
(lldb) image list
[ 0] /Volumes/work/gclayton/Documents/devb/attach/a.out
[ 1] /Developer/Platforms/iPhoneOS.platform/DeviceSupport/4.0/Symbols/usr/lib/dyld
[ 2] /Developer/Platforms/iPhoneOS.platform/DeviceSupport/4.0/Symbols/usr/lib/libSystem.B.dylib
Note that this is all happening prior to running _or_ connecting to a remote
platform. Once connected to a remote platform the OS version might change which
means we will need to update our dependecies. Also once we run, we will need
to match up the actualy binaries with the actualy UUID's to files in the
SDK, or download and cache them locally.
This is just the start of the remote platforms, but this modification is the
first iteration in getting the platforms really doing something.
llvm-svn: 127934
So the issue here was that we have lldb_private::FileSpec that by default was
always resolving a path when using the:
FileSpec::FileSpec (const char *path);
and in the:
void FileSpec::SetFile(const char *pathname, bool resolve = true);
This isn't what we want in many many cases. One example is you have "/tmp" on
your file system which is really "/private/tmp". You compile code in that
directory and end up with debug info that mentions "/tmp/file.c". Then you
type:
(lldb) breakpoint set --file file.c --line 5
If your current working directory is "/tmp", then "file.c" would be turned
into "/private/tmp/file.c" which won't match anything in the debug info.
Also, it should have been just a FileSpec with no directory and a filename
of "file.c" which could (and should) potentially match any instances of "file.c"
in the debug info.
So I removed the constructor that just takes a path:
FileSpec::FileSpec (const char *path); // REMOVED
You must now use the other constructor that has a "bool resolve" parameter that you must always supply:
FileSpec::FileSpec (const char *path, bool resolve);
I also removed the default parameter to SetFile():
void FileSpec::SetFile(const char *pathname, bool resolve);
And fixed all of the code to use the right settings.
llvm-svn: 116944
Arrange that this then gets properly set on attach, or when a "file" is set.
Add a completer for "process attach -n".
Caveats: there isn't currently a way to handle multiple processes with the same name. That
will have to wait on a way to pass annotations along with the completion strings.
llvm-svn: 110624
Greg Clayton