This reverts commit cf1c774d6a.
This change caused several regressions in the gdb test suite - at least
a sample of which was due to line zero instructions making breakpoints
un-lined. I think they're worth investigating/understanding more (&
possibly addressing) before moving forward with this change.
Revert "[FastISel] NFC: Clean up unnecessary bookkeeping"
This reverts commit 3fd39d3694.
Revert "[FastISel] NFC: Remove obsolete -fast-isel-sink-local-values option"
This reverts commit a474657e30.
Revert "Remove static function unused after cf1c774."
This reverts commit dc35368ccf.
Revert "[lldb] Fix TestThreadStepOut.py after "Flush local value map on every instruction""
This reverts commit 53a14a47ee.
Local values are constants or addresses that can't be folded into
the instruction that uses them. FastISel materializes these in a
"local value" area that always dominates the current insertion
point, to try to avoid materializing these values more than once
(per block).
https://reviews.llvm.org/D43093 added code to sink these local
value instructions to their first use, which has two beneficial
effects. One, it is likely to avoid some unnecessary spills and
reloads; two, it allows us to attach the debug location of the
user to the local value instruction. The latter effect can
improve the debugging experience for debuggers with a "set next
statement" feature, such as the Visual Studio debugger and PS4
debugger, because instructions to set up constants for a given
statement will be associated with the appropriate source line.
There are also some constants (primarily addresses) that could be
produced by no-op casts or GEP instructions; the main difference
from "local value" instructions is that these are values from
separate IR instructions, and therefore could have multiple users
across multiple basic blocks. D43093 avoided sinking these, even
though they were emitted to the same "local value" area as the
other instructions. The patch comment for D43093 states:
Local values may also be used by no-op casts, which adds the
register to the RegFixups table. Without reversing the RegFixups
map direction, we don't have enough information to sink these
instructions.
This patch undoes most of D43093, and instead flushes the local
value map after(*) every IR instruction, using that instruction's
debug location. This avoids sometimes incorrect locations used
previously, and emits instructions in a more natural order.
This does mean materialized values are not re-used across IR
instruction boundaries; however, only about 5% of those values
were reused in an experimental self-build of clang.
(*) Actually, just prior to the next instruction. It seems like
it would be cleaner the other way, but I was having trouble
getting that to work.
Differential Revision: https://reviews.llvm.org/D91734
I found a few cases where entries in the debug_line for a specific line of code have invalid entries (the address is outside of a code section or no section at all) and also valid entries. When this happens lldb might not set the breakpoint because the first line entry it will find in the line table might be the invalid one and since it's range is "invalid" no location is resolved. To get around this I changed the way we parse the line sequences to ignore those starting at an address under the first code segment.
Greg suggested to implement it this way so we don't need to check all sections for every line sequence.
Reviewed By: clayborg
Differential Revision: https://reviews.llvm.org/D87172
This would be reproducible in future DWZ category of the testsuite as:
Failed Tests (1):
lldb-api :: python_api/symbol-context/two-files/TestSymbolContextTwoFiles.py
Differential Revision: https://reviews.llvm.org/D91014
Current user_id_t format is:
63{isDebugTypes} 62..32{dwo || 7fffffff}
31..0 {die_offset}
while current DIERef format is (I have made up the bit positions but the
field widths do match):
63{m_section==isDebugTypes} 62{m_dwo_num_valid} 61..32{m_dwo_num}
31..0 {m_die_offset}
Proposing to change user_id_t to:
63{isDebugTypes} 62{dwo_is_valid} 61..32{dwo; 0 if !valid}
31..0 {die_offset}
There is no benefit of having 31-bits wide dwo_num in user_id_t when it
gets converted to 30-bits width in DIERef.
This patch is for future DWZ patchset which extends the dwo_is_valid bit
into a 2-bit field (normal, DWO, DWZ, DWZcommon) so that both user_id_t
and DIERef can be changed then the same way.
It would be best to somehow unify user_id_t and DIERef but I do not plan
to do that. user_id_t should probably remain a number for the Python API
compatibility while there still needs to be some class with all the
methods to access it.
SymbolFileDWARF::GetDwpSymbolFile() and SymbolFileDWARF::GetDIE use
0x3fffffff for DWP but that does not clash:
formerly:
31bits32..62:0x7fffffff = normal unit / not any DWO
31bits32..62:0x3fffffff = DWP
31bits32..62:others = DWO unit number
after this patch:
bit62=0 30bits32..61:any = normal unit / not any DWO
bit62=1 30bits32..61:0x3fffffff = DWP
bit62=1 30bits32..61:others = DWO unit number
Differential Revision: https://reviews.llvm.org/D90413
There were invalid DIE references which nobody used. If LLDB starts to
report invalid DIE references it would lock up (mutex lock).
These invalid DIE references are there since initial check-in by:
https://reviews.llvm.org/D83302
`image dump symtab` seems to output the symbols in whatever order they appear in
the DenseMap that is used to filter out symbols with non-unique addresses. As
DenseMap is a hash map this order can change at any time so the output of this
command is pretty unstable. This also causes the `Breakpad/symtab.test` to fail
with enabled reverse iteration (which reverses the DenseMap order to find issues
like this).
This patch makes the DenseMap a std::vector and uses a separate DenseSet to do
the address filtering. The output order is now dependent on the order in which
the symbols are read (which should be deterministic). It might also avoid a bit
of work as all the work for creating the Symbol constructor parameters is only
done when we can actually emplace a new Symbol.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D87036
The function was returning an incorrect (empty) value on the first
invocation. Given that this only affected the first invocation, this
bug/typo went mostly unaffected. DW_AT_const_value were particularly
badly affected by this as the GetByteSize call is
SymbolFileDWARF::ParseVariableDIE is likely to be the first call of this
function, and its effects cannot be undone by retrying.
Depends on D86348.
Differential Revision: https://reviews.llvm.org/D86436
Class-level static constexpr variables can have both DW_AT_const_value
(in the "declaration") and a DW_AT_location (in the "definition")
attributes. Our code was trying to handle this, but it was brittle and
hard to follow (and broken) because it was processing the attributes in
the order in which they were found.
Refactor the code to make the intent clearer -- DW_AT_location trumps
DW_AT_const_value, and fix the bug which meant that we were not
displaying these variables properly (the culprit was the delayed parsing
of the const_value attribute due to a need to fetch the variable type.
Differential Revision: https://reviews.llvm.org/D86615
This fixes several issues in handling of DW_AT_const_value attributes:
- the first is that the size of the data given by data forms does not
need to match the size of the underlying variable. We already had the
case to handle this for DW_FORM_(us)data -- this extends the handling
to other data forms. The main reason this was not picked up is because
clang uses leb forms in these cases while gcc prefers the fixed-size
ones.
- The handling of DW_AT_strp form was completely broken -- we would end
up using the pointer value as the result. I've reorganized this code
so that it handles all string forms uniformly.
- In case of a completely bogus form we would crash due to
strlen(nullptr).
Depends on D86311.
Differential Revision: https://reviews.llvm.org/D86348
Update the "image show-unwind" command output to show if the function
being shown is listed as a user-setting or platform trap handler.
Update the individual UnwindPlan dumps to show whether the unwind plan
is registered as a trap handler.
In some cases when we have a DW_AT_const_value and the data can be found in the
DWARFExpression then ValueObjectVariable does not handle it properly and we end
up with an extracting data from value failed error.
The test is a very stripped down assembly file since reproducing this relies on the results of compiling with -O1 which may not be stable over time.
Differential Revision: https://reviews.llvm.org/D86311
This is very similar to D85968, only more elusive to since we were not
adding the typedef type to the relevant DeclContext until D86140, which
meant that the DeclContext was populated (and the relevant assertion
hit) only after importing the type into the expression ast in a
particular way.
I haven't checked whether this situation can be hit in the gmodules
case, but my money is on "yes".
Differential Revision: https://reviews.llvm.org/D86216
Parsing DWARFv5 debug_loclist offsets when a CU is parsed is weighing
down memory usage of symbolizers that don't need to parse this data at
all. There's not much benefit to caching these anyway - since they are
O(1) lookup and reading once you know where the offset list starts (and
can do bounds checking with the offset list size too).
In general, I think it might be time to start paying down some of the
technical debt of loc/loclist/range/rnglist parsing to try to unify it a
bit more.
eg:
* Currently DWARFUnit has: RangeSection, RangeSectionBase, LocSection,
LocSectionBase, LocTable, RngListTable, LoclistTableHeader (be nice if
these were all wrapped up in two variables - one for loclists, one for
rnglists)
* rnglists and loclists are handled differently (see:
LoclistTableHeader, but no RnglistTableHeader)
* maybe all these types could be less stateful - lazily parse what they
need to, even reparsing rather than caching because it doesn't seem
too expensive, for instance. (though admittedly so long as it's
constantcost/overead per compilatiton that's probably adequate)
* Maybe implementing and using a DWARFDataExtractor that can be
sub-ranged (so we could slice it up to just the single contribution) -
though maybe that's not so useful because loc/ranges need to refer to
it by absolute, not contribution-relative mechanisms
Differential Revision: https://reviews.llvm.org/D86110
With -flimit-debug-info, we can run into cases when we only have a class
as a declaration, but we do have a definition of a nested class. In this
case, clang will hit an assertion when adding a member to an incomplete
type (but only if it's adding a c++ class, and not C struct).
It turns out we already had code to handle a similar situation arising
in the -gmodules scenario. This extends the code to handle
-flimit-debug-info as well, and reorganizes bits of other code handling
completion of types to move functions doing similar things closer
together.
Differential Revision: https://reviews.llvm.org/D85968
When bit-field data was stored in a Scalar in ValueObjectChild during UpdateValue()
it was extracting the bit-field value. Later on in lldb_private::DumpDataExtractor(…)
we were again attempting to extract the bit-field. Which would then not obtain the
correct value. This will remove the extra extraction in UpdateValue().
We hit this specific case when values are passed in registers, which we could only
reproduce in an optimized build.
Differential Revision: https://reviews.llvm.org/D85376
When loading a PE/COFF target, the associated PDB file often wasn't
found. The executable module contains a path for the associated PDB
file, but people often debug from a different directory than the one
their build system uses. (This is especially common in post-mortem
and cross platform debugging.)
Suppose the COFF executable being debugged is `~/proj/foo.exe`, but
it was built elsewhere and refers to `D:\remote\build\env\foobar.pdb`,
LLDB wouldn't find it.
With this change, if no file exists at the PDB path, LLDB will look
in the executable directory for a PDB file that matches the name of
the one it expected (e.g., `~/proj/foobar.pdb`). If found, the PDB
is subject to the same matching criteria (GUIDs and age) as would
have been used had it been in the original location.
This same-directory-as-the-binary rule is commonly used by debuggers
on Windows.
Differential Review: https://reviews.llvm.org/D84815
GNU ld allows sections after a non-SHF_ALLOC section to be covered by PT_LOAD
(PR37607) and assigns addresses to non-SHF_ALLOC output sections (similar to
SHF_ALLOC NOBITS sections. The location counter is not advanced).
This patch tries to fix PR37607 (remove a special case in
`Writer<ELFT>::createPhdrs`). To make the created PT_LOAD meaningful, we cannot
reset dot to 0 for a middle non-SHF_ALLOC output section. This results in
removal of two special cases in LinkerScript::assignOffsets. Non-SHF_ALLOC
non-orphan sections can have non-zero addresses like in GNU ld.
The zero address rule for non-SHF_ALLOC sections is weakened to apply to orphan
only. This results in a special case in createSection and findOrphanPos, respectively.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D85100
I have made the DW_FORM_ref4 relative. One could also use relocated
DW_FORM_ref_addr instead.
Tested with:
echo 'void f(){}'|clang -o 1.o -c -Wall -g -x c -;./bin/clang -o 1 1.o ../llvm-monorepo/lldb/test/Shell/SymbolFile/DWARF/DW_TAG_GNU_call_site-DW_AT_low_pc.s;./bin/lldb --no-lldbinit ./1 -o r -o 'p p' -o exit
Summary:
This effectively reverts r188124, which added code to handle
(DW_AT_)declarations of structures with some kinds of children as
definitions. The commit message claims this is a workaround for some
kind of debug info produced by gcc. However, it does not go into
specifics, so it's hard to reproduce or verify that this is indeed still a
problem.
Having this code is definitely a problem though, because it mistakenly
declares incomplete dwarf declarations to be complete. Both clang (with
-flimit-debug-info) and gcc (by default) generate DW_AT_declarations of
structs with children. This happens when full debug info for a class is
not emitted in a given compile unit (e.g. because of vtable homing), but
the class has inline methods which are used in the given compile unit.
In that case, the compilers emit a DW_AT_declaration of a class, but
add a DW_TAG_subprogram child to it to describe the inlined instance of
the method.
Even though the class tag has some children, it definitely does not
contain enough information to construct a full class definition (most
notably, it lacks any members). Keeping the class as incomplete allows
us to search for a real definition in other modules, helping the
-flimit-debug-info flow. And in case the definition is not found we can
display a error message saying that, instead of just showing an empty
struct.
Reviewers: clayborg, aprantl, JDevlieghere, shafik
Subscribers: lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D83302
Summary:
With D81784, lld has started debug info resolving relocations to
garbage-collected symbols as -1 (instead of relocation addend). For an
unaware consumer this generated sequences which seemingly wrap the
address space -- their first entry was 0xfffff, but all other entries
were low numbers.
Lldb stores line sequences concatenated into one large vector, sorted by
the first entry, and searched with std::lower_bound. This resulted in
the low-value entries being placed at the end of the vector, which
utterly confused the lower_bound algorithm, and caused it to not find a
match. (Previously, these sequences would be at the start of the vector,
and normally would contain addresses that are far smaller than any real
address we want to look up, so std::lower_bound was fine.)
This patch makes lldb ignore these kinds of sequences completely. It
does that by changing the construction algorithm from iterating over the
rows (as parsed by llvm), to iterating over the sequences. This is
important because the llvm parsed performs validity checks when
constructing the sequence array, whereas the row array contains raw
data.
Reviewers: JDevlieghere, MaskRay
Differential Revision: https://reviews.llvm.org/D83957
Summary:
-debug-info-kind=constructor reduces the amount of class debug info that
is emitted; this patch switches to using this as the default.
Constructor homing emits the complete type info for a class only when the
constructor is emitted, so it is expected that there will be some classes that
are not defined in the debug info anymore because they are never constructed,
and we shouldn't need debug info for these classes.
I compared the PDB files for clang, and there are 273 class types that are defined with `=limited`
but not with `=constructor` (out of ~60,000 total class types).
We've looked at a number of the types that are no longer defined with =constructor. The vast
majority of cases are something like class A is used as a parameter in a member function of
some other class B, which is emitted. But the function that uses class A is never called, and class A
is never constructed, and therefore isn't emitted in the debug info.
Bug: https://bugs.llvm.org/show_bug.cgi?id=46537
Subscribers: aprantl, cfe-commits, lldb-commits
Tags: #clang, #lldb
Differential Revision: https://reviews.llvm.org/D79147
With -flimit-debug-info, we can have a definition of a class, but no
definition for some of its members. This extends the same logic we were
using for incomplete base classes to cover incomplete members too.
Test forward-declarations.s is removed as it is no longer applicable --
we don't warn anymore when encountering incomplete members as they could
be completed elsewhere. New checks added to TestLimitDebugInfo cover the
handling of incomplete members more thoroughly.
This complements the existing TestLimitDebugInfo.py, which tests this
scenario more comprehensively, but is not able to run on all hosts.
Specifically, it's hard to trigger this code from windows because clang
tries hard to ensure that debug info for types marked with
__declspec(dllexport) is emitted even under -flimit-debug-info (and
dllexport is needed to use a type across shared libraries).
This assembly-based test serves two purposes:
- it tests that -flimit-debug-info code path works for windows binaries
(even though the aforementioned feature means its less likely to be
used there)
- it gives basic test coverage for the -flimit-debug-info handling code
when running the test suite on windows hosts.
Summary:
When evaluating an expression referencing a constexpr static member variable, an
error is issued because the PDB does not specify a symbol with an address that
can be relocated against.
Rather than attempt to resolve the variable's value within the IR execution, the
values of all constants can be looked up and incorporated into the AST of the
record type as a literal, mirroring the original compiler AST.
This change applies to DIA and native PDB loaders.
Patch By: jackoalan
Reviewers: aleksandr.urakov, jasonmolenda, zturner, jdoerfert, teemperor
Reviewed By: aleksandr.urakov
Subscribers: sstefan1, lldb-commits, llvm-commits, #lldb
Tags: #lldb, #llvm
Differential Revision: https://reviews.llvm.org/D82160
D80519 <https://reviews.llvm.org/D80519>
added support for `DW_TAG_GNU_call_site` but
Bug 45886 <https://bugs.llvm.org/show_bug.cgi?id=45886>
found one case did not work.
There is:
0x000000b1: DW_TAG_GNU_call_site
DW_AT_low_pc (0x000000000040111e)
DW_AT_abstract_origin (0x000000cc "a")
...
0x000000cc: DW_TAG_subprogram
DW_AT_name ("a")
DW_AT_prototyped (true)
DW_AT_low_pc (0x0000000000401109)
^^^^^^^^^^^^ - here it did overwrite the 'low_pc' variable containing value 0x40111e we wanted
DW_AT_high_pc (0x0000000000401114)
DW_AT_frame_base (DW_OP_call_frame_cfa)
DW_AT_GNU_all_call_sites (true)
DW_TAG_GNU_call_site attributes order as produced by GCC:
0x000000b1: DW_TAG_GNU_call_site
DW_AT_low_pc (0x000000000040111e)
DW_AT_abstract_origin (0x000000cc "a")
clang produces the attributes in opposite order:
0x00000064: DW_TAG_GNU_call_site
DW_AT_abstract_origin (0x0000002a "a")
DW_AT_low_pc (0x0000000000401146)
Differential Revision: https://reviews.llvm.org/D81334
Previously, we were simply ignoring them and continuing the evaluation.
This behavior does not seem useful, because the resulting value will
most likely be completely bogus.
Summary:
For ObjCInterfaceDecls, LLDB iterates over the `methods` of the interface in FindExternalVisibleDeclsByName
since commit ef423a3ba5 .
However, when LLDB calls `oid->methods()` in that function, Clang will pull in all declarations in the current
DeclContext from the current ExternalASTSource (which is again, `ClangExternalASTSourceCallbacks`). The
reason for that is that `methods()` is just a wrapper for `decls()` which is supposed to provide a list of *all*
(both currently loaded and external) decls in the DeclContext.
However, `ClangExternalASTSourceCallbacks::FindExternalLexicalDecls` doesn't implement support for ObjCInterfaceDecl,
so we don't actually add any declarations and just mark the ObjCInterfaceDecl as having no ExternalLexicalStorage.
As LLDB uses the ExternalLexicalStorage to see if it can complete a type with the ExternalASTSource, this causes
that LLDB thinks our class can't be completed any further by the ExternalASTSource
and will from on no longer make any CompleteType/FindExternalLexicalDecls calls to that decl. This essentially
renders those types unusable in the expression parser as they will always be considered incomplete.
This patch just changes the call to `methods` (which is just a `decls()` wrapper), to some ad-hoc `noload_methods`
call which is wrapping `noload_decls()`. `noload_decls()` won't trigger any calls to the ExternalASTSource, so
this prevents that ExternalLexicalStorage will be set to false.
The test for this is just adding a method to an ObjC interface. Before this patch, this unset the ExternalLexicalStorage
flag and put the interface into the state described above.
In a normal user session this situation was triggered by setting a breakpoint in a method of some ObjC class. This
caused LLDB to create the MethodDecl for that specific method and put it into the the ObjCInterfaceDecl.
Also `ObjCLanguageRuntime::LookupInCompleteClassCache` needs to be unable to resolve the type do
an actual definition when the breakpoint is set (I'm not sure how exactly this can happen, but we just
found no Type instance that had the `TypePayloadClang::IsCompleteObjCClass` flag set in its payload in
the situation where this happens. This however doesn't seem to be a regression as logic wasn't changed
from what I can see).
The module-ownership.mm test had to be changed as the only reason why the ObjC interface in that test had
it's ExternalLexicalStorage flag set to false was because of this unintended side effect. What actually happens
in the test is that ExternalLexicalStorage is first set to false in `DWARFASTParserClang::CompleteTypeFromDWARF`
when we try to complete the `SomeClass` interface, but is then the flag is set back to true once we add
the last ivar of `SomeClass` (see `SetMemberOwningModule` in `TypeSystemClang.cpp` which is called
when we add the ivar). I'll fix the code for that in a follow-up patch.
I think some of the code here needs some rethinking. LLDB and Clang shouldn't infer anything about the ExternalASTSource
and its ability to complete the current type form the `ExternalLexicalStorage` flag. We probably should
also actually provide any declarations when we get asked for the lexical decls of an ObjCInterfaceDecl. But both of those
changes are bigger (and most likely would cause us to eagerly complete more types), so those will be follow up patches
and this patch just brings us back to the state before commit ef423a3ba5 .
Fixes rdar://63584164
Reviewers: aprantl, friss, shafik
Reviewed By: aprantl, shafik
Subscribers: arphaman, abidh, JDevlieghere
Differential Revision: https://reviews.llvm.org/D80556
Muhammad Omair Javaid