expression evaluation.
OrcMCJITReplacement is a reimplementation of MCJIT using ORC components, and
provides an easy upgrade path to ORC for existing MCJIT clients. There should be
no functional changes resulting from this switch.
llvm-svn: 279327
Options used to store a reference to the CommandInterpreter instance
in the base Options class. This made it impossible to parse options
independent of a CommandInterpreter.
This change removes the reference from the base class. Instead, it
modifies the options-parsing-related methods to take an
ExecutionContext pointer, which the options may inspect if they need
to do so.
Closes https://reviews.llvm.org/D23416
Reviewers: clayborg, jingham
llvm-svn: 278440
This feature was added to solve a lookup problem in expressions when local variables
shadow ivars. That solution requires fully realizing all local variables to evaluate
any expression, and can cause significant performance problems when evaluating
expressions in frames that have many complex locals.
Until we get a better solution, this setting mitigates the problem when you don't
have local variables that shadow ivars.
<rdar://problem/27226122>
llvm-svn: 274783
We had support that assumed that thread local data for a variable could be determined solely from the module in which the variable exists. While this work for linux, it doesn't work for Apple OSs. The DWARF for thread local variables consists of location opcodes that do something like:
DW_OP_const8u (x)
DW_OP_form_tls_address
or
DW_OP_const8u (x)
DW_OP_GNU_push_tls_address
The "x" is allowed to be anything that is needed to determine the location of the variable. For Linux "x" is the offset within the TLS data for a given executable (ModuleSP in LLDB). For Apple OS variants, it is the file address of the data structure that contains a pthread key that can be used with pthread_getspecific() and the offset needed.
This fix passes the "x" along to the thread:
virtual lldb::addr_t
lldb_private::Thread::GetThreadLocalData(const lldb::ModuleSP module, lldb::addr_t tls_file_addr);
Then this is passed along to the DynamicLoader::GetThreadLocalData():
virtual lldb::addr_t
lldb_private::DynamicLoader::GetThreadLocalData(const lldb::ModuleSP module, const lldb::ThreadSP thread, lldb::addr_t tls_file_addr);
This allows each DynamicLoader plug-in do the right thing for the current OS.
The DynamicLoaderMacOSXDYLD was modified to be able to grab the pthread key from the data structure that is in memory and call "void *pthread_getspecific(pthread_key_t key)" to get the value of the thread local storage and it caches it per thread since it never changes.
I had to update the test case to access the thread local data before trying to print it as on Apple OS variants, thread locals are not available unless they have been accessed at least one by the current thread.
I also added a new lldb::ValueType named "eValueTypeVariableThreadLocal" so that we can ask SBValue objects for their ValueType and be able to tell when we have a thread local variable.
<rdar://problem/23308080>
llvm-svn: 274366
Previously we eliminated the randomized scheme for finding memory when the
underlying process cannot allocate memory, and replaced it with an algorithm
that starts the allocations at 00x.
This was more determinstic, but runs into problems on embedded targets where the
pages near 0x0 are in fact interesting memory. To deal with those cases, this
patch does two things:
- It makes the default fallback be an address that is less likely than 0x0 to
contain interesting information.
- Before falling back to this, it adds an algorithm that consults the
GetMemoryRegionInfo() API to see if it can find an unmapped area.
This should eliminate the randomness (and unpredictable memory accesseas) of the
previous scheme while making expressions more likely to return correct results.
<rdar://problem/25545573>
llvm-svn: 272301
Summary:
Without this commit, when `log enable lldb expr` is enabled, the
disassembly of JIT'ed code is never displayed.
Reviewers: spyffe, clayborg
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D20312
llvm-svn: 271863
Some compilers do not mark up C++ functions as extern "C" in the DWARF, so LLDB
has to fall back (if it is about to give up finding a symbol) to using the base
name of the function.
This fix also ensures that we search by full name rather than "auto," which
could cause unrelated C++ names to be found. Finally, it adds a test case.
<rdar://problem/25094302>
llvm-svn: 271551
This is a pretty straightforward first pass over removing a number of uses of
Mutex in favor of std::mutex or std::recursive_mutex. The problem is that there
are interfaces which take Mutex::Locker & to lock internal locks. This patch
cleans up most of the easy cases. The only non-trivial change is in
CommandObjectTarget.cpp where a Mutex::Locker was split into two.
llvm-svn: 269877
This allows expressions such as 'i == 1 || i == 2` to be executed using the IR interpreter, instead of relying on JIT code injection (which may not be available on some platforms).
Patch by cameron314
Differential Revision: http://reviews.llvm.org/D19124
llvm-svn: 269340
1. Fixed semicolon placement in the lambda in the test itself.
2. Fixed lldbinline tests in general so that we don't attempt tests on platforms that don't use the given type of debug info. (For example, no DWO tests on Windows.) This fixes one of the two failures on Windows. (TestLambdas.py was the only inline test that wasn't XFailed or skipped on Windows.)
3. Set the error string in IRInterpreter::CanInterpret so that the caller doesn't print (null) instead of an explanation. I don't entirely understand the error, so feel free to suggest a better wording.
4. XFailed the test on Windows. The interpreter won't evaluate the lambda because the module has multiple function bodies. I don't exactly understand why that's a problem for the interpreter nor why the problem arises only on Windows.
Differential Revision: http://reviews.llvm.org/D19606
llvm-svn: 268573
In templated const functions, trying to run an expression would produce the
error
error: out-of-line definition of '$__lldb_expr' does not match any declaration
in 'foo' member declaration does not match because it is const qualified
error: 1 error parsing expression
which is no good. It turned out we don't actually need to worry about "const,"
we just need to be consistent about the declaration of the expression and the
FunctionDecl we inject into the class for "this."
Also added a test case.
<rdar://problem/24985958>
llvm-svn: 268083
This reverts commit r267833 as it breaks the build. It looks like some work in progress got
committed together with the actual fix, but I'm not sure which one is which, so I'll revert the
whole patch and let author resumbit it after fixing the build error.
llvm-svn: 267861
In templated const functions, trying to run an expression would produce the
error
error: out-of-line definition of '$__lldb_expr' does not match any declaration in 'foo'
member declaration does not match because it is const qualified
error: 1 error parsing expression
which is no good. It turned out we don't actually need to worry about "const,"
we just need to be consistent about the declaration of the expression and the
FunctionDecl we inject into the class for "this."
Also added a test case.
<rdar://problem/24985958>
llvm-svn: 267833
but when there's was no process it was just returning an null pointer
and not setting the error. I don't have a scenario where this might
go wrong, just code inspection...
llvm-svn: 267594
statements for, be sure not to include variables that have no locations. We wouldn't
be able to realize them, and that will cause all expressions to fail.
llvm-svn: 267500
Recommit modified version of r266311 including build bot regression fix.
This differs from the original r266311 by:
- Fixing Scalar::Promote to correctly zero- or sign-extend value depending
on signedness of the *source* type, not the target type.
- Omitting a few stand-alone fixes that were already committed separately.
llvm-svn: 266422
The Scalar implementation and a few other places in LLDB directly
access the internal implementation of APInt values using the
getRawData method. Unfortunately, pretty much all of these places
do not handle big-endian systems correctly. While on little-endian
machines, the pointer returned by getRawData can simply be used as
a pointer to the integer value in its natural format, no matter
what size, this is not true on big-endian systems: getRawData
actually points to an array of type uint64_t, with the first element
of the array always containing the least-significant word of the
integer. This means that if the bitsize of that integer is smaller
than 64, we need to add an offset to the pointer returned by
getRawData in order to access the value in its natural type, and
if the bitsize is *larger* than 64, we actually have to swap the
constituent words before we can access the value in its natural type.
This patch fixes every incorrect use of getRawData in the code base.
For the most part, this is done by simply removing uses of getRawData
in the first place, and using other APInt member functions to operate
on the integer data.
This can be done in many member functions of Scalar itself, as well
as in Symbol/Type.h and in IRInterpreter::Interpret. For the latter,
I've had to add a Scalar::MakeUnsigned routine to parallel the existing
Scalar::MakeSigned, e.g. in order to implement an unsigned divide.
The Scalar::RawUInt, Scalar::RawULong, and Scalar::RawULongLong
were already unused and can be simply removed. I've also removed
the Scalar::GetRawBits64 function and its few users.
The one remaining user of getRawData in Scalar.cpp is GetBytes.
I've implemented all the cases described above to correctly
implement access to the underlying integer data on big-endian
systems. GetData now simply calls GetBytes instead of reimplementing
its contents.
Finally, two places in the clang interface code were also accessing
APInt.getRawData in order to actually construct a byte representation
of an integer. I've changed those to make use of a Scalar instead,
to avoid having to re-implement the logic there.
The patch also adds a couple of unit tests verifying correct operation
of the GetBytes routine as well as the conversion routines. Those tests
actually exposed more problems in the Scalar code: the SetValueFromData
routine didn't work correctly for 128- and 256-bit data types, and the
SChar routine should have an explicit "signed char" return type to work
correctly on platforms where char defaults to unsigned.
Differential Revision: http://reviews.llvm.org/D18981
llvm-svn: 266311
The result variables aren't useful, and if you have a breakpoint on a
common function you can generate a lot of these. So I changed the
code that checks the condition to set ResultVariableIsInternal in the
EvaluateExpressionOptions that we pass to the execution.
Unfortunately, the check for this variable was done in the wrong place
(the static UserExpression::Evaluate) which is not how breakpoint
conditions execute expressions (UserExpression::Execute). So I moved
the check to UserExpression::Execute (which Evaluate also calls) and made the
overridden method DoExecute.
llvm-svn: 266093
They're not supposed to go in the symbol table, and in fact the way the JIT
is currently implemented it sometimes crashes when you try to get the
address of such a function. So we skip them.
llvm-svn: 264821
quietly apply fixits for those who really trust clang's fixits.
Also, moved the retry into ClangUserExpression::Evaluate, where I can make a whole new ClangUserExpression
to do the work. Reusing any of the parts of a UserExpression in situ isn't supported at present.
<rdar://problem/25351938>
llvm-svn: 264793
This feature is controlled by an expression command option, a target property and the
SBExpressionOptions setting. FixIt's are only applied to UserExpressions, not UtilityFunctions,
those you have to get right when you make them.
This is just a first stage. At present the fixits are applied silently. The next step
is to tell the user about the applied fixit.
<rdar://problem/25351938>
llvm-svn: 264379
IRExecutionUnits contain code and data that persistent declarations can
depend on. In order to keep them alive and provide for lookup of these
symbols, we now allow any PersistentExpressionState to keep a list of
execution units. Then, when doing symbol lookup on behalf of an
expression, any IRExecutionUnit can consult the persistent expression
states on a particular Target to find the appropriate symbol.
<rdar://problem/22864976>
llvm-svn: 263995
a way for compilation to take a "thread to use for compilation". If it isn't set then the
compilation will use the currently selected thread. This should help keep function execution
to the one thread intended.
llvm-svn: 263972
We want to do a better job presenting errors that occur when evaluating
expressions. Key to this effort is getting away from a model where all
errors are spat out onto a stream where the client has to take or leave
all of them.
To this end, this patch adds a new class, DiagnosticManager, which
contains errors produced by the compiler or by LLDB as an expression
is created. The DiagnosticManager can dump itself to a log as well as
to a string. Clients will (in the future) be able to filter out the
errors they're interested in by ID or present subsets of these errors
to the user.
This patch is not intended to change the *users* of errors - only to
thread DiagnosticManagers to all the places where streams are used. I
also attempt to standardize our use of errors a bit, removing trailing
newlines and making clients omit 'error:', 'warning:' etc. and instead
pass the Severity flag.
The patch is testsuite-neutral, with modifications to one part of the
MI tests because it relied on "error: error:" being erroneously
printed. This patch fixes the MI variable handling and the testcase.
<rdar://problem/22864976>
llvm-svn: 263859
Summary:
From Adrian McCarthy:
"Running ninja check-lldb now has one crash in a Python process, due to deferencing a null pointer in IRExecutionUnit.cpp: candidate_sc.symbol is null, which leads to a call with a null this pointer."
Reviewers: zturner, spyffe, amccarth
Subscribers: ted, jingham, lldb-commits
Differential Revision: http://reviews.llvm.org/D17860
llvm-svn: 263066
The System-V x86_64 ABI requires floating point values to be passed
in 128-but SSE vector registers (xmm0, ...). When printing such a
variable this currently yields an <invalid load address>.
This patch makes LLDB's DWARF expression evaluator accept 128-bit
registers as scalars. It also relaxes the check that the size of the
result of the DWARF expression be equal to the size of the variable to a
greater-than. DWARF defers to the ABI how smaller values are being placed
in a larger register.
Implementation note: I found the code in Value::SetContext() that changes
the m_value_type after the fact to be questionable. I added a sanity check
that the Value's memory buffer has indeed been written to (this is
necessary, because we may have a scalar value in a vector register), but
really I feel like this is the wrong place to be setting it.
Reviewed by Greg Clayton.
http://reviews.llvm.org/D17897
rdar://problem/24944340
llvm-svn: 262947
Summary: Recent changes to the expression parser broke function name resolution when using the IR interpreter instead of JIT. This patch changes the IRMemoryMap ivar in InterpreterStackFrame to an IRExecutionUnitSP (which is a subclass), allowing InterpreterStackFrame::ResolveConstantValue() to call FindSymbol() on the name of the Value when it's a FunctionVal. It also changes IRExecutionUnit::FindInSymbols() to call GetFileAddress() on the symball if ResolveCallableAddress() fails and there is no valid Process.
Reviewers: spyffe
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D17745
llvm-svn: 262407
If we have a TargetLoadAddress on the top of the DWARF stack then a
DW_OP_plus or a DW_OP_plus_ucons sholudn't dereference (resolve) it
and then add the value to the dereferenced value but it should offset
the load address by the specified constant.
llvm-svn: 262339
The evaluation of expressions containing register values was broken for targets for which endianness differs from host.
Committed on behalf of: mamai <marianne.mailhot.sarrasin@gmail.com>
Differential revision: http://reviews.llvm.org/D17167
llvm-svn: 262041
Additionally fix the type of some dwarf expression where we had a
confusion between scalar and load address types after a dereference.
Differential revision: http://reviews.llvm.org/D17604
llvm-svn: 262014
IRExecutionUnit previously replicated a bunch of logic that already
existed elsewhere for the purpose of getting a load address for a
symbol. This approach failed to resolve certain types of symbols.
Instead, we now use functions on SymbolContext to do the address
resolution.
This is a cleanup of IRExecutionUnit::FindInSymbols, and also fixes a
latent bug where we looked at the wrong SymbolContext to determine
whether or not it is external.
<rdar://problem/24770829>
llvm-svn: 261704
the xcode project file to catch switch statements that have a
case that falls through unintentionally.
Define LLVM_FALLTHROUGH to indicate instances where a case has code
and intends to fall through. This should be in llvm/Support/Compiler.h;
Peter Collingbourne originally checked in there (r237766), then
reverted (r237941) because he didn't have time to mark up all the
'case' statements that were intended to fall through. I put together
a patch to get this back in llvm http://reviews.llvm.org/D17063 but
it hasn't been approved in the past week. I added a new
lldb-private-defines.h to hold the definition for now.
Every place in lldb where there is a comment that the fall-through
is intentional, I added LLVM_FALLTHROUGH to silence the warning.
I haven't tried to identify whether the fallthrough is a bug or
not in the other places.
I haven't tried to add this to the cmake option build flags.
This warning will only work for clang.
This build cleanly (with some new warnings) on macosx with clang
under xcodebuild, but if this causes problems for people on other
configurations, I'll back it out.
llvm-svn: 260930
Since IRExecutionUnit is now capable of looking up symbols, and the JIT is up to
the task of generating the appropriate relocations, we don't need to do all the
work that IRForTarget used to do to fixup symbols at the IR level.
We also don't need to allocate data manually (with its attendant bugs) because
the JIT is capable of doing so without crashing.
We also don't need the awkward lldb.call.realName metadata to determine what
calls are objc_msgSend, because they now just reference objc_msgSend.
To make this work, we ensure that we recognize which symbols are extern "C" and
report them to the compiler as such. We also report the full Decl of functions
rather than just making up top-level functions with the appropriate types.
This should not break any testcases, but let me know if you run into any issues.
<rdar://problem/22864926>
llvm-svn: 260768
Previously we would try both versions of a symbol -- the one with _ in it and
the one without -- in all cases, because we didn't know what the current
platform's policy was. However, stripping _ is only necessary on platforms
where _ is the prefix for global symbols.
There's an API that does this, though, on llvm::DataLayout, so this patch fixes
IRExecutionUnit to use that API to determine whether or not to strip _ from the
symbol or not.
llvm-svn: 260767
If an instruction has a constant that IRInterpreter doesn't know how to deal
with (say, an array constant, because we can't materialize it to APInt) then we
used to ignore that and only fail during expression execution. This is annoying
because if IRInterpreter had just returned false from CanInterpret(), the JIT
would have been used.
Now the IRInterpreter checks constants as part of CanInterpret(), so this should
hopefully no longer be an issue.
llvm-svn: 260735
I'm preparing to remove symbol lookup from IRForTarget, where it constitutes a
dreadful hack working around no-longer-existing JIT bugs. Thanks to our
contributors, IRForTarget has a lot of smarts that IRExecutionUnit doesn't have,
so I've cleaned them up a bit and moved them over to IRExecutionUnit.
Also for historical reasons, IRExecutionUnit used the "Small" code model on non-
ELF platforms (namely, OS X). That's no longer necessary, and we can use the
same code model as everyone else on OS X. I've fixed that.
llvm-svn: 260734
Lang Hames