Summary:
This patch attempts to solve two issues made this code hard to follow
for me.
The first issue was that a lot of what these visitors do is mutate the
AST. The visitor pattern is not particularly good for that because by
the time you have performed the dynamic type dispatch, it's too late to
go back to the parent node, and change its pointer. The previous code
dealt with that relatively elegantly, but it still meant that one had to
perform manual type checks, which is what the visitor pattern is
supposed to avoid.
The second issue was not being able to return values from the Visit
functions, which meant that one had to store function results in member
variables (a common problem with visitor patterns).
Here, I solve both problems by making the visitor use a type switch
instead of going through double dispatch on the visited object. This
allows one to parameterize the visitor based on the return type and pass
function results as function results. The mutation is fascilitated by
having each Visit function take two arguments -- a reference to the
object itself (with the correct dynamic type), and a reference to the
parent's pointer to this object.
Although this wasn't my explicit goal here, the fact that we're not
using virtual dispatch anymore allows us to make the AST nodes
trivially destructible, which is a good thing, since we were not
destroying them anyway.
Reviewers: aleksandr.urakov, amccarth
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D60410
llvm-svn: 358261
A lot of comments in LLDB are surrounded by an ASCII line to delimit the
begging and end of the comment.
Its use is not really consistent across the code base, sometimes the
lines are longer, sometimes they are shorter and sometimes they are
omitted. Furthermore, it looks kind of weird with the 80 column limit,
where the comment actually extends past the line, but not by much.
Furthermore, when /// is used for Doxygen comments, it looks
particularly odd. And when // is used, it incorrectly gives the
impression that it's actually a Doxygen comment.
I assume these lines were added to improve distinguishing between
comments and code. However, given that todays editors and IDEs do a
great job at highlighting comments, I think it's worth to drop this for
the sake of consistency. The alternative is fixing all the
inconsistencies, which would create a lot more churn.
Differential revision: https://reviews.llvm.org/D60508
llvm-svn: 358135
Summary:
The code was passing pointers around, expecting they would be not null.
In c++ it is possible to convey this notion explicitly by using a
reference instead.
Not all uses of pointers could be converted to references (e.g. one
can't store references in a container), but this will at least make it
locally obvious that code is dealing with nonnull pointers.
Reviewers: aleksandr.urakov, amccarth
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D60271
llvm-svn: 357744
Summary:
Now CVType and CVSymbol are effectively type-safe wrappers around
ArrayRef<uint8_t>. Make the kind() accessor load it from the
RecordPrefix, which is the same for types and symbols.
Reviewers: zturner, aganea
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D60018
llvm-svn: 357658
A recent patch to LLD started emitting information about import modules.
These are represented as compile units in the PDB, but with no
additional debug info. This was confusing the native pdb reader, who
expected that the debug info stream be present.
This should fix failing tests on the Windows bots.
llvm-svn: 357513
Summary:
This refactors moves the register name->number resolution out of the
FPOProgramNodeRegisterRef class. Instead I create a special
FPOProgramNodeSymbol class, which holds unresolved symbols, and move the
resolution into the ResolveRegisterRefs visitor.
The background here is that I'd like to use this code for Breakpad
unwind info, which uses similar syntax to describe unwind info. For
example, a simple breakpad unwind program might look like:
.cfa: $esp 8 + $ebp: .cfa 8 - ^
To be able to do this, I need to be able to customize register
resolving, as that is presently hardcoded to use codeview register
names, but breakpad supports a lot more architectures with different
register names. Moving the resolution into a separate class will allow
each user to use a different resolution logic.
Reviewers: aleksandr.urakov, zturner, amccarth
Subscribers: jdoerfert, lldb-commits
Differential Revision: https://reviews.llvm.org/D60068
llvm-svn: 357455
ICF can cause multiple symbols to start at the same virtual address.
I plan to handle this shortly, but I wanted to correct the comment for
now.
Deleted an obsolete comment about adjusting the offset for the magic
number at the beginning of the debug info stream. This adjustment is
handled at a lower level now.
llvm-svn: 355943
My apologies for the large patch. With the exception of ConstString.h
itself it was entirely produced by sed.
ConstString has exactly one const char * data member, so passing a
ConstString by reference is not any more efficient than copying it by
value. In both cases a single pointer is passed. But passing it by
value makes it harder to accidentally return the address of a local
object.
(This fixes rdar://problem/48640859 for the Apple folks)
Differential Revision: https://reviews.llvm.org/D59030
llvm-svn: 355553
COFF files are modelled in lldb as having one big container section
spanning the entire module image, with the actual sections being
subsections of that. In this model, the base address is simply the
address of the first byte of that section.
This also removes the hack where ObjectFilePECOFF was using the
m_file_offset field to communicate this information. Using file offset
for this purpose is completely wrong, as that is supposed to indicate
where is this ObjectFile located in the file on disk. This field is only
meaningful for fat binaries, and should normally be 0.
Both PDB plugins have been updated to use GetBaseAddress instead of
GetFileOffset.
llvm-svn: 354258
This patch properly extracts the full submodule path as well as its
search paths from DWARF import decls and passes it on to the
ClangModulesDeclVendor.
rdar://problem/47970144
Differential Revision: https://reviews.llvm.org/D58090
llvm-svn: 353961
Summary:
This is coming from the discussion in D55356 (the most interesting part
happened on the mailing list, so it isn't reflected on the review page).
In short the issue is that lldb assumes that all bytes of a module image
in memory will be backed by a "section". This isn't the case for PECOFF
files because the initial bytes of the module image will contain the
file header, which does not correspond to any normal section in the
file. In particular, this means it is not possible to implement
GetBaseAddress function for PECOFF files, because that's supposed point
to the first byte of that header.
If my (limited) understanding of how PECOFF files work is correct, then
the OS is expecded to load the entire module into one continuous chunk
of memory. The address of that chunk (+/- ASLR) is given by the "image
base" field in the COFF header, and it's size by "image size". All of
the COFF sections are then loaded into this range.
If that's true, then we can model this behavior in lldb by creating a
"container" section to represent the entire module image, and then place
other sections inside that. This would make be consistent with how MachO
and ELF files are modelled (except that those can have multiple
top-level containers as they can be loaded into multiple discontinuous
chunks of memory).
This change required a small number of fixups in the PDB plugins, which
assumed a certain order of sections within the object file (which
obivously changes now). I fix this by changing the lookup code to use
section IDs (which are unchanged) instead of indexes. This has the nice
benefit of removing spurious -1s in the plugins as the section IDs in
the pdbs match the 1-based section IDs in the COFF plugin.
Besides making the implementation of GetBaseAddress possible, this also
improves the lookup of addresses in the gaps between the object file
sections, which will now be correctly resolved as belonging to the
object file.
Reviewers: zturner, amccarth, stella.stamenova, clayborg, lemo
Reviewed By: clayborg, lemo
Subscribers: JDevlieghere, abidh, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D56537
llvm-svn: 353916
Summary:
This patch makes virtual bases to be added in the correct order to the bases
list. It is important because `VTableContext` (`MicrosoftVTableContext` in our
case) uses then the order of virtual bases in the list to restore the virtual
table indexes. These indexes are used then to resolve the layout of the virtual
bases.
We haven't enough information about offsets of virtual bases regarding to the
object (moreover, in a common case we can't rely on such information, see the
example here: https://reviews.llvm.org/D53506#1272306 ), but there should be
enough information to restore the layout of the virtual bases from the indexes
in runtime. After D53506 this information is used whenever possible, so there
should be no problems with virtual bases' fields reading.
Reviewers: zturner, rnk, stella.stamenova
Subscribers: abidh, teemperor, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D56904
llvm-svn: 353806
stored relative to VFRAME
Summary:
This patch makes LLDB able to retrieve proper values for function arguments and
local variables stored in PDB relative to VFRAME register.
Patch contains retrieval of corresponding FPO table entries from PDB and a
generic translator from FPO programs to DWARF expressions to get correct VFRAME
value.
Patch also improves variables-locations.test and makes this test passable on
x86.
Patch By: leonid.mashinsky
Reviewers: zturner, asmith, stella.stamenova, aleksandr.urakov
Reviewed By: zturner
Subscribers: arphaman, labath, mgorny, aprantl, JDevlieghere, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D55122
llvm-svn: 352845
Summary:
This patch adds the basic support of methods reconstruction by native PDB
plugin. It contains only most obvious changes (it processes LF_ONEMETHOD and
LF_METHOD records), some things still remain unsolved:
- mangled names retrieving;
- support of template methods.
Reviewers: zturner, labath, lemo, stella.stamenova
Reviewed by: zturner
Differential Revision: https://reviews.llvm.org/D56126
llvm-svn: 352464
to reflect the new license.
We understand that people may be surprised that we're moving the header
entirely to discuss the new license. We checked this carefully with the
Foundation's lawyer and we believe this is the correct approach.
Essentially, all code in the project is now made available by the LLVM
project under our new license, so you will see that the license headers
include that license only. Some of our contributors have contributed
code under our old license, and accordingly, we have retained a copy of
our old license notice in the top-level files in each project and
repository.
llvm-svn: 351636
This parameter was only ever used with the Module set, and
since a SymbolFile is tied to a module, the parameter turns
out to be entirely unnecessary. Furthermore, it doesn't make
a lot of sense to ask a caller to ask SymbolFile which is tied
to Module X to find types for Module Y, but that possibility
was open with the previous interface. By removing this
parameter from the API, it makes it harder to use incorrectly
as well as easier for an implementor to understand what it
needs to do.
llvm-svn: 351133
Every callsite was passing an empty SymbolContext, so this parameter
had no effect. Inside the DWARF implementation of this function,
however, there was one codepath that checked members of the
SymbolContext. Since no call-sites actually ever used this
functionality, it was essentially dead code, so I've deleted this
code path as well.
llvm-svn: 351132
This method took a SymbolContext but only actually cared about the
case where the m_function member was set. Furthermore, it was
intended to be implemented to parse blocks recursively despite not
documenting this in its name. So we change the name to indicate
that it should be recursive, while also limiting the function
parameter to be a Function&. This lets the caller know what is
required to use it, as well as letting new implementers know what
kind of inputs they need to be prepared to handle.
llvm-svn: 351131
Previously all of these functions accepted a SymbolContext&.
While a CompileUnit is one member of a SymbolContext, there
are also many others, and by passing such a monolithic parameter
in this way it makes the requirements and assumptions of the
API unclear for both callers as well as implementors.
All these methods need is a CompileUnit. By limiting the
parameter type in this way, we simplify the code as well as
make it self-documenting for both implementers and users.
Differential Revision: https://reviews.llvm.org/D56564
llvm-svn: 350943
The function SymbolFile::ParseTypes previously accepted a SymbolContext.
This makes it extremely difficult to implement faithfully, because you
have to account for all possible combinations of members being set in
the SymbolContext. On the other hand, no clients of this function
actually care about implementing this function to this strict of a
standard. AFAICT, there is actually only 1 client in the entire
codebase, and it is the function ParseAllDebugSymbols, which is itself
only called for testing purposes when dumping information. At this
call-site, the only field it sets is the CompileUnit, meaning that an
implementer of a SymbolFile need not worry about any examining or
handling any other fields which might be set.
By restricting this API to accept exactly a CompileUnit& and nothing
more, we can simplify the life of new SymbolFile plugin implementers by
making it clear exactly what the necessary and sufficient set of
functionality they need to implement is, while at the same time removing
some dead code that tried to handle other types of SymbolContext fields
that were never going to be set anyway.
Differential Revision: https://reviews.llvm.org/D56462
llvm-svn: 350889
Typedefs are represented as S_UDT records in the globals stream. This
creates a strange situation where "types" are actually represented as
"symbols", so they need special handling.
In order to test this, we don't just use lldb and print out some
variables causing the AST to get created, because variables whose type
is a typedef will have debug info referencing the original type, not the
typedef. So we use lldb-test instead which will parse all debug info in
the entire file. This exposed some problems with lldb-test and the
native reader, mainly that certain types of obscure symbols which we can
find when iterating every single record would trigger crashes. These
have been fixed as well so that lldb-test can be used to test this
functionality.
Differential Revision: https://reviews.llvm.org/D56461
llvm-svn: 350888
This is a first step towards getting lldb-test symbols working
with the native plugin. There is a remaining issue, which is
that the plugin expects that ParseDeclsForContext will also
create lldb symbols rather than just the decls, but the native
pdb plugin doesn't currently do this. This will be addressed
in a followup patch.
llvm-svn: 350243
There were several problems preventing this from working. The
first is that when the PDB had an absolute path to the main
source file, we would construct an invalid path by prepending the
compilation directory to it anyway. So we needed to check if the
path is already absolute first.
Second, LLDB assumes that the zero'th item in the support file list
is the main compilation unit. We were respecting this requirement,
but LLDB *also* requires that file to appear somewhere in the list
starting from index 1 as well. So the main compilation file should
appear in the support file list twice. And when parsing a line
table, it expects the LineEntry records to be constructed using
the 1-based index. With these two fixes we can now set breakpoints
by file and line using the native PDB reader.
llvm-svn: 350240
Previously we would create these for local variables but not for
global variables.
Also updated existing tests which created global variables to check
for them in the resulting AST.
llvm-svn: 349854
We had a use after free where we were assigning the result of a function
that returned a string to a StringRef. After fixing this use after
free, one of the DIA PDB tests now passes with the native PDB reader,
so we enable the test under native mode as well. The goal is to
eventually make all the tests pass under both, at which point we can
disable them all under DIA mode.
llvm-svn: 349673
We reconstruct the AST hierarchy by trying to hack up a mangled
name for the parent type using the child type's mangled name.
This was failing for enums because their tag type is represented
with two letters ("W4") instead of one letter ("T", "U", etc) as
it is with classes, structs, and unions. After accounting for
this we can now correctly determine when an enum is nested
inside of a namespace or a class.
llvm-svn: 349565
Previously the code that parsed debug info to create lldb's Symbol
objects such as Variable, Type, Function, etc was tightly coupled
to the AST reconstruction code. This made it difficult / impossible
to implement functions such as ParseDeclsForContext() that were only
supposed to be operating on clang AST's. By splitting these apart,
the logic becomes much cleaner and we have a clear separation of
responsibilities.
llvm-svn: 349383
This patch adds support for parsing and evaluating local variables.
using the native pdb plugin.
Differential Revision: https://reviews.llvm.org/D55575
llvm-svn: 349067
Previously CreateParameterDeclaration was always using the translation
unit DeclContext. We would later go and add parameters to the
FunctionDecl, but internally clang makes a copy when you do this, and
we'd end up with ParmVarDecl's at the global scope as well as in the
function scope.
This fixes the issue. It's hard to say whether this will introduce
a behavioral change in name lookup, but I know there have been several
hacks introduced in previous years to deal with collisions between
various types of variables, so there's a chance that this patch could
obviate one of those hacks.
Differential Revision: https://reviews.llvm.org/D55571
llvm-svn: 348941
Previously we would create an lldb::Function object for each function
parsed, but we would not add these to the clang AST. This is a first
step towards getting local variable support working, as we first need an
AST decl so that when we create local variable entries, they have the
proper DeclContext.
Differential Revision: https://reviews.llvm.org/D55384
llvm-svn: 348631
VarStreamArray was built on the assumption that it is backed by a
StreamRef, and offset 0 of that StreamRef is the first byte of the first
record in the array.
This is a logical and intuitive assumption, but unfortunately we have
use cases where it doesn't hold. Specifically, a PDB module's symbol
stream is prefixed by 4 bytes containing a magic value, and the first
byte of record data in the array is actually at offset 4 of this byte
sequence.
Previously, we would just truncate the first 4 bytes and then construct
the VarStreamArray with the resulting StreamRef, so that offset 0 of the
underlying stream did correspond to the first byte of the first record,
but this is problematic, because symbol records reference other symbol
records by the absolute offset including that initial magic 4 bytes. So
if another record wants to refer to the first record in the array, it
would say "the record at offset 4".
This led to extremely confusing hacks and semantics in loading code, and
after spending 30 minutes trying to get some math right and failing, I
decided to fix this in the underlying implementation of VarStreamArray.
Now, we can say that a stream is skewed by a particular amount. This
way, when we access a record by absolute offset, we can use the same
values that the records themselves contain, instead of having to do
fixups.
Differential Revision: https://reviews.llvm.org/D55344
llvm-svn: 348499
Originally we created our 64-bit UID scheme by using the first byte as
sort of a "tag" to represent what kind of symbol this was, and we
re-used the PDB_SymType enumeration for this. For native pdb support,
this is not really the right abstraction layer, because what we really
want is something that tells us *how* to find the symbol. This means,
specifically, is in the globals stream / public stream / module stream /
TPI stream / etc, and for whichever one it is in, where is it within
that stream?
A good example of why the old namespacing scheme was insufficient is
that it is more or less impossible to create a uid for a field list
member of a class/struction/union/enum that tells you how to locate
the original record.
With this new scheme, the first byte is no longer a PDB_SymType enum
but a new enum created specifically to identify where in the PDB
this record lives. This gives us much better flexibility in
what kinds of symbols the uids can identify.
llvm-svn: 347018
clang-cl does not emit these, but MSVC does, so we need to be able to
handle them.
Because clang-cl does not generate them, it was a bit hard to write a
test. So what I had to do was get an PDB file with some S_CONSTANT
records in using cl and link, dump it using llvm-pdbutil dump -globals
-sym-data to get the bytes of the records, generate the same object file
using clang-cl but with -S to emit an assembly file, and replace all the
S_LDATA32 records with the bytes of the S_CONSTANT records. This way, we
can compile the file using llvm-mc and link it with lld-link.
Differential Revision: https://reviews.llvm.org/D54452
llvm-svn: 346787
In a previous patch, we pre-processed the TPI stream in order to build
the reverse mapping from nested type -> parent type so that we could
accurately reconstruct a DeclContext hierarchy.
However, there were some issues. An LF_NESTTYPE record is really just a
typedef, so although it happens to be used to indicate the name of the
nested type and referring to the global record which defines the type,
it is also used for every other kind of nested typedef. When we rebuild
the DeclContext hierarchy, we want it to be as accurate as possible,
which means that if we have something like:
struct A {
struct B {};
using C = B;
};
We don't want to create two CXXRecordDecls in the AST each with the
exact same definition. We just want to create one for B and then
define C as an alias to B. Previously, however, it would not be able
to distinguish between the two cases and it would treat A::B and
A::C as being two classes each with separate definitions. We address
the first half of improving the pre-processing logic so that only
actual definitions are treated this way.
Later, in a followup patch, we can handle the case of nested
typedefs since we're already going to be enumerating the field list
anyway and this patch introduces the general framework for
distinguishing between the two cases.
Differential Revision: https://reviews.llvm.org/D54357
llvm-svn: 346786
This was originally submitted in a patch which fixed two unrelated
bugs at the same time. This portion of the fix was reverted because
it broke several other things. However, the fix employed originally
was totally wrong, and attempted to change something in the ValueObject
printer when actually the bug was in the NativePDB plugin. We need
to mark forward enum decls as having external storage, otherwise
we won't be asked to complete them when the time comes. This patch
implements the proper fix, and updates tests accordingly.
llvm-svn: 346517
Bitfields are represented as LF_MEMBER records whose TypeIndex
points to an LF_BITFIELD record that describes the bit width,
bit offset, and underlying type of the bitfield. All we need to
do is resolve these when resolving record types.
llvm-svn: 346511
Moved the declaration of m_kind below the declaration of cvclass,
cvunion and cvenum. This order is necessary because in one of the
constructors the initialization of m_kind depends on the value of
cvclass.
third_party/llvm/llvm/tools/lldb/source/Plugins/SymbolFile/NativePDB/PdbUtil.cpp:50:7: error: field 'cvclass' will be initialized after field 'm_kind' [-Werror,-Wreorder]
: cvclass(std::move(c)),
^
third_party/llvm/llvm/tools/lldb/source/Plugins/SymbolFile/NativePDB/PdbUtil.cpp:51:14: error: field 'cvclass' is uninitialized when used here [-Werror,-Wuninitialized]
m_kind(cvclass.Kind == TypeRecordKind::Struct ? Struct : Class) {}
llvm-svn: 346435
