When the type of a function parameter sees its top CV qualifier
change, that should never negatively affect ABI compliance.
So this patch filters out top cv qualifier changes on function
parameter types, by default.
* include/abg-comparison.h (enum diff_category): Add a new
FN_PARM_TYPE_TOP_CV_CHANGE_CATEGORY enumerator. "Or" the
enumerator to the EVERYTHING_CATEGORY enumerator.
* src/abg-comp-filter.cc (has_fn_parm_type_cv_qual_change): Define
new static function.
(categorize_harmless_diff_node): Categorize changes to top cv
qualifiers on function parameter types into the new
FN_PARM_TYPE_TOP_CV_CHANGE_CATEGORY.
* src/abg-comparison.cc (get_default_harmless_categories_bitmap):
Add the new FN_PARM_TYPE_TOP_CV_CHANGE_CATEGORY category to the
set of harmless categories.
(operator<<(ostream&, diff_category)): Adjust to serialize
the new FN_PARM_TYPE_TOP_CV_CHANGE_CATEGORY.
* tests/data/test-diff-filter/libtest40-v0.so: New test input binary.
* tests/data/test-diff-filter/libtest40-v1.so: Likewise.
* tests/data/test-diff-filter/test40-report-0.txt: New test
reference output.
* tests/data/test-diff-filter/test40-v0.cc: Source code of the
test binary above.
* tests/data/test-diff-filter/test40-v1.cc: Likewise.
* tests/data/Makefile.am: Add the new test material above to
source distribution.
* tests/test-diff-filter.cc (in_out_specs): Add new binaries to
compare.
* tests/data/test-diff-filter/test30-pr18904-rvalueref-report0.txt:
Adjust.
* tests/data/test-diff-filter/test30-pr18904-rvalueref-report1.txt:
Likewise.
* tests/data/test-diff-filter/test35-pr18754-no-added-syms-report-0.txt:
Likewise.
* tests/data/test-diff-filter/test35-pr18754-no-added-syms-report-1.txt:
Likewise.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Sometimes some virtual member functions don't have any virtual offset
set in the debug info. This happens for virtual destructors
sometimes. In that case, the ABI change report should not refer to
that unset virtual offset as being '0'. Rather, it shouldn't refer to
it at all.
This is what this patch does.
* include/abg-ir.h (mem_fn_context_rel::mem_fn_context_rel):
Initialize the virtual offset to -1.
* src/abg-comparison.cc (represent): In the overload to represent
a method_decl, do not represent the vofffset if it's not set.
* src/abg-writer.cc (write_voffset): The virtual offset is signed
because if it's -1, it means no offset is set.
* tests/data/test-annotate/test14-pr18893.so.abi: Adjust.
* tests/data/test-annotate/test18-pr19037-libvtkRenderingLIC-6.1.so.abi:
Adjust.
* tests/data/test-annotate/test19-pr19023-libtcmalloc_and_profiler.so.abi:
Adjust.
* tests/data/test-annotate/test20-pr19025-libvtkParallelCore-6.1.so.abi:
Adjust.
* tests/data/test-diff-dwarf-abixml/test0-pr19026-libvtkIOSQL-6.1.so.1.abi:
Adjust.
* tests/data/test-diff-dwarf/test28-vtable-changes-report-0.txt: Adjust.
* tests/data/test-read-dwarf/test10-pr18818-gcc.so.abi: Adjust.
* tests/data/test-read-dwarf/test11-pr18828.so.abi: Adjust.
* tests/data/test-read-dwarf/test14-pr18893.so.abi: Adjust.
* tests/data/test-read-dwarf/test18-pr19037-libvtkRenderingLIC-6.1.so.abi:
Adjust.
* tests/data/test-read-dwarf/test19-pr19023-libtcmalloc_and_profiler.so.abi:
Adjust.
* tests/data/test-read-dwarf/test20-pr19025-libvtkParallelCore-6.1.so.abi:
Adjust.
* tests/data/test-read-dwarf/test22-pr19097-libstdc++.so.6.0.17.so.abi:
Adjust.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
When a class is forward-declared, resolving it to a definition that
appears later in the same translation unit or in another translation
is an interesting problem.
Until now, the declaration would be resolved to the definition of that
class found in the binary. The problem is that there can be different
such definitions, especially in C where there is no "One Definition
Rule". In that case, the definition chosen is random.
This patch resolves that randomness.
For a given class declaration, if there is just one possible
definition in the binary, then the declaration is resolved to that
definition. If there is one definition for that declaration in the
same translation unit, then the declaration is resolved to that
definition. If there are more than one definitions in translation
units that are not the one of the declaration, then the declaration is
left unresolved. This is what I call "selective class declaration resolution".
Note that an unresolved class declaration now compares different to a
definition of a class of the same name. This is so that we can have
an unresolved class be present in the resulting .abi file, alongside
an (incompatible) definition of the same class. The change from a class
declaration to its definition is filtered out by default, though.
* include/abg-fwd.h (type_base_wptrs_type)
(istring_type_base_wptrs_map_type): Define new typedefs.
(lookup_class_types): Declare new functions.
* include/abg-ir.h
(environment::decl_only_class_equals_definition): Declare new
accessor.
(type_maps::{*_types}): Make these accessors return
istring_type_base_wptrs_map_type& instead of
istring_type_base_wptr_map_type&.
* src/abg-dwarf-reader.cc
(read_context::resolve_declaration_only_classes): Implement the
new selective declaration resolution scheme.
* src/abg-ir.cc (type_maps::priv::{*_types_}): Change the type of
these data members from istring_type_base_wptr_map_type to
istring_type_base_wptrs_map_type.
(type_maps::{*_types}): Make these accessors definitions return
istring_type_base_wptrs_map_type& instead of
istring_type_base_wptr_map_type&.
(translation_unit::bind_function_type_life_time): Adjust.
(environment::priv::decl_only_class_equals_definition_): New data
member.
(environment::priv::priv): Initialize it. By default, a decl-only
class is now considered different from its definition.
(environment::decl_only_class_equals_definition): Define new
accessor.
(lookup_types_in_map, lookup_class_types): Define new functions.
(lookup_type_in_map, lookup_union_type_per_location)
(lookup_basic_type, lookup_basic_type_per_location)
(lookup_class_type, lookup_class_type_per_location)
(lookup_union_type, lookup_enum_type)
(lookup_enum_type_per_location, lookup_typedef_type)
(lookup_typedef_type_per_location, lookup_qualified_type)
(lookup_pointer_type, lookup_reference_type, lookup_array_type)
(lookup_function_type, maybe_update_types_lookup_map)
(maybe_update_types_lookup_map<class_decl>)
(maybe_update_types_lookup_map<function_type>): Adjust.
(type_base::get_canonical_type_for): When doing type comparison
here, we can now consider that an unresolved class declaration
compares different to an incompatible class definition of the same
name. So no need to look through decl-only classes in that case.
(equals): In the overload for class_or_union, if
environment::decl_only_class_equals_definition() is false, then an
unresolved class declaration of name "N" compares different to a
class definition named "N".
* tests/data/test-annotate/test15-pr18892.so.abi: Adjust.
* tests/data/test-read-dwarf/test9-pr18818-clang.so.abi: Adjust
* tests/data/test-read-dwarf/test12-pr18844.so.abi: Adjust.
* tests/data/test-read-dwarf/test15-pr18892.so.abi: Adjust.
* tests/data/test-diff-dwarf/test28-vtable-changes-report-0.txt:
Adjust.
* tests/data/test-diff-pkg/spice-server-0.12.4-19.el7.x86_64-0.12.8-1.el7.x86_64-report-2.txt:
Adjust.
* tests/data/test-diff-filter/test38/Makefile: New test material.
* tests/data/test-diff-filter/test38/test38-a.c: Likewise.
* tests/data/test-diff-filter/test38/test38-b.c: Likewise.
* tests/data/test-diff-filter/test38/test38-c.c: Likewise.
* tests/data/test-diff-filter/test38/test38-report-0.txt: Likewise.
* tests/data/test-diff-filter/test38/test38-v0: Likewise.
* tests/data/test-diff-filter/test38/test38-v1: Likewise.
* tests/data/test-diff-filter/test38/test38.h: Likewise.
* tests/data/test-diff-filter/test39/Makefile: Likewise.
* tests/data/test-diff-filter/test39/test39-a-v0.c: Likewise.
* tests/data/test-diff-filter/test39/test39-a-v1.c: Likewise.
* tests/data/test-diff-filter/test39/test39-b-v0.c: Likewise.
* tests/data/test-diff-filter/test39/test39-b-v1.c: Likewise.
* tests/data/test-diff-filter/test39/test39-c-v0.c: Likewise.
* tests/data/test-diff-filter/test39/test39-c-v1.c: Likewise.
* tests/data/test-diff-filter/test39/test39-main.c: Likewise.
* tests/data/test-diff-filter/test39/test39-report-0.txt: Likewise.
* tests/data/test-diff-filter/test39/test39-v0: Likewise.
* tests/data/test-diff-filter/test39/test39-v1: Likewise.
* tests/data/test-diff-filter/test39/test39.h: Likewise.
* tests/data/Makefile.am: Add the new test material above to the
source distribution.
* tests/test-diff-filter.cc (in_out_specs): Add the new test
inputs above to the test harness.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Right now, when loading each corpus of a group, a new read_context is
created and destroyed. That makes thousands of corpora that are
created and destroyed. Profiling seems to argue that we'd gain in
performance by re-using the first read_context that was created
instead, and re-set it before loading a new corpus.
This is what this patch does.
* include/abg-dwarf-reader.h (reset_read_context): Declare new
function.
* src/abg-dwarf-reader.cc (read_context::elf_paths_): Make this to
be non const.
(read_context::initialize): New function to initialize all data
members.
(read_context::read_context): Use the new read_context::initialize
function, rather than initializing data members 'inline' here.
(reset_read_context): Define a new function to reset a
read_context so that it can be re-used to load a new corpus.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
When using abidw to generate an abixml for a Linux Kernel build tree,
usually, people have to copy the vmlinux binary into the directory
where modules are, so that the tool can find it. This --vmlinux
option helps to avoid doing that copy.
Simarly, when comparing two Linux Kernel build trees, --vmlinux1 and
--vmlinux2 are there to make the tool find the vmlinux binaries to
compare, independantly from the directories under which the modules
are to be found.
* include/abg-tools-utils.h
(build_corpus_group_from_kernel_dist_under): Add a new
vmlinux_path parameter.
* src/abg-tools-utils.cc (find_vmlinux_and_module_paths): Do not
try to find a vmlinux binary if we already have the path to one.
(build_corpus_group_from_kernel_dist_under): Add a new
vmlinux_path parameter.
* tools/abidw.cc (options::vmlinux): New data member.
(display_usage): Add a usage string for --vmlinux
(parse_command_line): Parse the new --vmlinux option.
(load_kernel_corpus_group_and_write_abixml): Fix some return code
when the function fails. Verify the presence of the vmlinux
binary that was given. Adjust.
* tools/kmidiff.cc (options::{vmlinux1, vmlinux2}): New data
members.
(display_usage): Add a usage string for --vmlinux1 and --vmlinux2.
(parse_command_line): Parse the --vmlinux1 and --vmlinux2
options.
(main): Adjust.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
get_type_name can be called over and over again on function types.
This patch caches the result of the computation of function type names
in those cases.
This is a speed optimization.
* src/abg-ir.cc (get_type_name): Cache function type names.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
* src/abg-reader.cc (build_enum_type_decl): Do not check for
errno which might have been set earlier by something else.
Rather, check the returned value for overflow or underflow.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Sometimes some virtual member functions don't have any virtual offset
set in the debug info. This happens for virtual destructors
sometimes. In that case, the ABI change report should not refer to
that unset virtual offset as being '0'. Rather, it shouldn't refer to
it at all.
This is what this patch does.
* include/abg-ir.h (mem_fn_context_rel::mem_fn_context_rel):
Initialize the virtual offset to -1.
* src/abg-comparison.cc (represent): In the overload to represent
a method_decl, do not represent the vofffset if it's not set.
* src/abg-writer.cc (write_voffset): The virtual offset is signed
because if it's -1, it means no offset is set.
* tests/data/test-annotate/test14-pr18893.so.abi: Adjust.
* tests/data/test-annotate/test18-pr19037-libvtkRenderingLIC-6.1.so.abi:
Adjust.
* tests/data/test-annotate/test19-pr19023-libtcmalloc_and_profiler.so.abi:
Adjust.
* tests/data/test-annotate/test20-pr19025-libvtkParallelCore-6.1.so.abi:
Adjust.
* tests/data/test-diff-dwarf-abixml/test0-pr19026-libvtkIOSQL-6.1.so.1.abi:
Adjust.
* tests/data/test-diff-dwarf/test28-vtable-changes-report-0.txt: Adjust.
* tests/data/test-read-dwarf/test10-pr18818-gcc.so.abi: Adjust.
* tests/data/test-read-dwarf/test11-pr18828.so.abi: Adjust.
* tests/data/test-read-dwarf/test14-pr18893.so.abi: Adjust.
* tests/data/test-read-dwarf/test18-pr19037-libvtkRenderingLIC-6.1.so.abi:
Adjust.
* tests/data/test-read-dwarf/test19-pr19023-libtcmalloc_and_profiler.so.abi:
Adjust.
* tests/data/test-read-dwarf/test20-pr19025-libvtkParallelCore-6.1.so.abi:
Adjust.
* tests/data/test-read-dwarf/test22-pr19097-libstdc++.so.6.0.17.so.abi:
Adjust.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
For DIE originating from C, we now compute canonical DIEs. We then
use that to compare DIEs to see if they are equal or not. So string
representation of DIEs are now used only to reduce the number of DIEs
comparisons that is performed during DIE canonicalization.
We can thus just use function names (rather than a full
die_function_signature) as a way to reduce the number of structural
comparisons of DIEs during canonicalization.
This patch does just that.
Note that in the future when we perform DIEs canonicalization and comparison for
C++, we can avoid computing full function DIE signatures for C++ too.
* src/abg-dwarf-reader.cc (die_function_signature): For C DIEs,
just return the (linkage) name of the function.
* tests/data/test-annotate/test15-pr18892.so.abi: Adjust.
* tests/data/test-annotate/test21-pr19092.so.abi: Adjust.
* tests/data/test-read-dwarf/test15-pr18892.so.abi: Adjust.
* tests/data/test-read-dwarf/test21-pr19092.so.abi: Adjust.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
When a class is forward-declared, resolving it to a definition that
appears later in the same translation unit or in another translation
is an interesting problem.
Until now, the declaration would be resolved to the definition of that
class found in the binary. The problem is that there can be different
such definitions, especially in C where there is no "One Definition
Rule". In that case, the definition chosen is random.
This patch resolves that randomness.
For a given class declaration, if there is just one possible
definition in the binary, then the declaration is resolved to that
definition. If there is one definition for that declaration in the
same translation unit, then the declaration is resolved to that
definition. If there are more than one definitions in translation
units that are not the one of the declaration, then the declaration is
left unresolved. This is what I call "selective class declaration resolution".
Note that an unresolved class declaration now compares different to a
definition of a class of the same name. This is so that we can have
an unresolved class be present in the resulting .abi file, alongside
an (incompatible) definition of the same class. The change from a class
declaration to its definition is filtered out by default, though.
* include/abg-fwd.h (type_base_wptrs_type)
(istring_type_base_wptrs_map_type): Define new typedefs.
(lookup_class_types): Declare new functions.
* include/abg-ir.h
(environment::decl_only_class_equals_definition): Declare new
accessor.
(type_maps::{*_types}): Make these accessors return
istring_type_base_wptrs_map_type& instead of
istring_type_base_wptr_map_type&.
* src/abg-dwarf-reader.cc
(read_context::resolve_declaration_only_classes): Implement the
new selective declaration resolution scheme.
* src/abg-ir.cc (type_maps::priv::{*_types_}): Change the type of
these data members from istring_type_base_wptr_map_type to
istring_type_base_wptrs_map_type.
(type_maps::{*_types}): Make these accessors definitions return
istring_type_base_wptrs_map_type& instead of
istring_type_base_wptr_map_type&.
(translation_unit::bind_function_type_life_time): Adjust.
(environment::priv::decl_only_class_equals_definition_): New data
member.
(environment::priv::priv): Initialize it. By default, a decl-only
class is now considered different from its definition.
(environment::decl_only_class_equals_definition): Define new
accessor.
(lookup_types_in_map, lookup_class_types): Define new functions.
(lookup_type_in_map, lookup_union_type_per_location)
(lookup_basic_type, lookup_basic_type_per_location)
(lookup_class_type, lookup_class_type_per_location)
(lookup_union_type, lookup_enum_type)
(lookup_enum_type_per_location, lookup_typedef_type)
(lookup_typedef_type_per_location, lookup_qualified_type)
(lookup_pointer_type, lookup_reference_type, lookup_array_type)
(lookup_function_type, maybe_update_types_lookup_map)
(maybe_update_types_lookup_map<class_decl>)
(maybe_update_types_lookup_map<function_type>): Adjust.
(type_base::get_canonical_type_for): When doing type comparison
here, we can now consider that an unresolved class declaration
compares different to an incompatible class definition of the same
name. So no need to look through decl-only classes in that case.
(equals): In the overload for class_or_union, if
environment::decl_only_class_equals_definition() is false, then an
unresolved class declaration of name "N" compares different to a
class definition named "N".
* tests/data/test-annotate/test15-pr18892.so.abi: Adjust.
* tests/data/test-read-dwarf/test9-pr18818-clang.so.abi: Adjust
* tests/data/test-read-dwarf/test12-pr18844.so.abi: Adjust.
* tests/data/test-read-dwarf/test15-pr18892.so.abi: Adjust.
* tests/data/test-diff-dwarf/test28-vtable-changes-report-0.txt:
Adjust.
* tests/data/test-diff-pkg/spice-server-0.12.4-19.el7.x86_64-0.12.8-1.el7.x86_64-report-2.txt:
Adjust.
* tests/data/test-diff-filter/test38/Makefile: New test material.
* tests/data/test-diff-filter/test38/test38-a.c: Likewise.
* tests/data/test-diff-filter/test38/test38-b.c: Likewise.
* tests/data/test-diff-filter/test38/test38-c.c: Likewise.
* tests/data/test-diff-filter/test38/test38-report-0.txt: Likewise.
* tests/data/test-diff-filter/test38/test38-v0: Likewise.
* tests/data/test-diff-filter/test38/test38-v1: Likewise.
* tests/data/test-diff-filter/test38/test38.h: Likewise.
* tests/data/test-diff-filter/test39/Makefile: Likewise.
* tests/data/test-diff-filter/test39/test39-a-v0.c: Likewise.
* tests/data/test-diff-filter/test39/test39-a-v1.c: Likewise.
* tests/data/test-diff-filter/test39/test39-b-v0.c: Likewise.
* tests/data/test-diff-filter/test39/test39-b-v1.c: Likewise.
* tests/data/test-diff-filter/test39/test39-c-v0.c: Likewise.
* tests/data/test-diff-filter/test39/test39-c-v1.c: Likewise.
* tests/data/test-diff-filter/test39/test39-main.c: Likewise.
* tests/data/test-diff-filter/test39/test39-report-0.txt: Likewise.
* tests/data/test-diff-filter/test39/test39-v0: Likewise.
* tests/data/test-diff-filter/test39/test39-v1: Likewise.
* tests/data/test-diff-filter/test39/test39.h: Likewise.
* tests/data/Makefile.am: Add the new test material above to the
source distribution.
* tests/test-diff-filter.cc (in_out_specs): Add the new test
inputs above to the test harness.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
This patch introduces the concept of canonical DIEs.
To date, when two C types have the same string representation [1] and
are defined at the same location, they are considered to be the same,
just as in C++.
[1]: String representation of a type: For a type named
"Foo" that is a structure, the representation is "struct Foo"
This patch introduces more finesse in determining if two types are
different.
It implements comparing the DIEs of the types, directly from the DWARF
representation. When comparing pointers, typedefs and qualified
types, the underlying type is compared recursively.
The type de-duplication scheme is now centered around two data
structures.
1/ A map that associates the string representation of a type with
a vector of the offsets of the type DIEs that have the same
representation. Members of this vector denotes DIEs of types that
are *different* even if they all have the same representation.
Each DIE in a given vector is the canonical DIE of its class of
equivalence. This map is the map of all canonical DIEs organized
by the representation of those canonical DIEs.
2/ A map that associates the offset of the canonical type DIE with
the resulting internal representation of the type. Here, the
internal representation is an instance of the
abigail::ir::type_base type. This map is the map of the types
associated to the canonical type DIEs.
There is also a vector that associates a DIE 'D' to its canonical DIE
'C'. The index of the vector is the offset of 'D' and the value of
the element at that index is 'C'.
Thus, each time we are about to create (or get) an internal
representation for a type DIE denoted 'D', we first get the canonical
DIE of 'D', denoted C. If C doesn't exist, we create it. That is, we
add a new entry in the map 1/. Then we look in map 2/ to see if the C
(and thus D) has a associated type.
If C has an associated type, we return it.
If C has no associated type, we create a type for it (and thus for D)
and we associate the new type to the offsets of both D and C.
The rest of the patch is mostly boiler plate and adjustment to
accomodate this new de-duplication scheme.
* src/abg-dwarf-reader.cc (die_decl_map_type, die_type_map_type):
Remove these typedefs.
(die_artefact_map_type, istring_dwarf_offsets_map_type): New
typedefs.
(die_is_at_class_scope, die_qualified_type_name)
(die_qualified_decl_name, die_qualified_type_name_empty)
(die_return_and_parm_names_from_fn_type_die)
(die_function_type_is_method_type):
Const-ify the read_context& parameter.
(read_context::die_source_dependant_container_set::get_container):
Likewise.
(read_context::{name_artefacts_map_, per_tu_name_artefacts_map_,
die_decl_map_, alternate_die_decl_map_, type_unit_die_decl_map_,
die_type_map_, alternate_die_type_map_, type_unit_die_type_map_}):
Remove data members.
(read_context::{die_decl_map, alternate_die_decl_map,
associate_die_to_decl_primary, associate_die_to_decl_alternate,
associate_die_to_decl_from_type_unit,
lookup_decl_from_die_offset_primary,
lookup_decl_from_die_offset_alternate,
lookup_decl_from_type_unit_die_offset,
lookup_type_artifact_from_die_per_tu,
lookup_artifact_from_per_tu_die_representation,
associate_die_to_artifact_by_repr,
associate_die_to_artifact_by_repr_internal, clear_die_type_maps}):
Remove member functions.
(read_context::{decl_die_repr_die_offsets_maps_,
type_die_repr_die_offsets_maps_, decl_die_artefact_maps_,
type_die_artefact_maps_, dwarf_expr_eval_context_}): Add new data
members.
(read_context::clear_per_translation_unit_data): Don't clear
read_context::per_tu_name_artefacts_map_ data member as it's
removed.
(read_context::clear_per_corpus_data): Don't clear
read_context::name_artefacts_map_ and all the other relevant data
members that got removed.
(read_context::{dwarf_per_die_source,
decl_die_repr_die_offsets_maps, type_die_repr_die_offsets_maps,
get_canonical_die, get_die_from_offset, decl_die_artefact_maps,
type_die_artefact_maps, dwarf_expr_eval_ctxt}): Add new member
functions.
(compare_dies, compare_as_decl_dies)
(compare_as_type_dies, maybe_finish_function_decl_reading)
(die_is_anonymous): Define new functions.
(read_context::associate_die_to_decl): Remove the
do_associate_by_repr_per_tu parameter. Use the new
read_context::{decl_die_artefact_maps_, get_canonical_die} member
functions.
(read_context::lookup_decl_from_die_offset): Use Dwarf_Off rather
than size_t for the type of the die_offset parameter. Use the
lookup_artifact_from_die_offset member function.
(read_context::lookup_type_artifact_from_die): Const-ify. In one
overload, take a new 'die_as_type' parameter. Use the new
get_canonical_die, type_die_artefact_maps and
decl_die_artefact_maps member functions. In the second overload,
use the first overload.
(read_context::odr_is_relevant): Add an overload that takes a DIE.
(read_context::associate_die_to_type): Remove the
do_associate_by_repr and do_associate_per_tu parameters. Use the
new get_canonical_die and type_die_artefact_maps member functions.
(read_context::lookup_type_from_die): Use the new
lookup_artifact_from_die member function.
(read_context::lookup_type_from_die_offset): Use the new
type_die_artefact_maps member function. When the found artifact
is a function_decl, return its type.
(read_context::schedule_type_for_late_canonicalization): Use the
new get_canonical_die and type_die_artefact_maps member functions.
(die_function_signature): Const-ify. Get the scope name right
even for scopes that are not types.
(die_member_offset): Make eval_last_constant_dwarf_sub_expr use
the new cached DWARF expression evalution context.
(get_parent_die): Support where_offset equals to zero. This means
we are looking at a C binary, basically.
(build_enum_type) : Use the new overload of
read_context::odr_is_relevant that takes a DIE. Adjust.
(add_or_update_union_type, add_or_update_class_type): Don't lookup
classes/unions per location anymore. Now that we can compare DIEs
in a fined grain manner, the approximation of the location is not
useful anymore.
(build_pointer_type)
(build_function_type): Associate DIE to type if we reuse an
existing type.
(build_or_get_fn_decl_if_not_suppressed): When re-using a
function decl internal representation from an equivalent DIE that
we've seen before, it can happen that we want to augment that
function decl internal representation with new properties coming
from the DIE we are currently looking at; do that here.
(is_function_for_die_a_member_of_class): Remove the "where_offset"
parameter.
(add_or_update_member_function): Adjust.
* tests/data/test-annotate/libtest23.so.abi: Adjust.
* tests/data/test-annotate/test13-pr18894.so.abi: Adjust.
* tests/data/test-annotate/test14-pr18893.so.abi: Adjust.
* tests/data/test-annotate/test15-pr18892.so.abi: Adjust.
* tests/data/test-annotate/test17-pr19027.so.abi: Adjust.
* tests/data/test-annotate/test18-pr19037-libvtkRenderingLIC-6.1.so.abi: Adjust.
* tests/data/test-annotate/test19-pr19023-libtcmalloc_and_profiler.so.abi: Adjust.
* tests/data/test-annotate/test20-pr19025-libvtkParallelCore-6.1.so.abi: Adjust.
* tests/data/test-annotate/test21-pr19092.so.abi: Adjust.
* tests/data/test-diff-pkg/libICE-1.0.6-1.el6.x86_64.rpm--libICE-1.0.9-2.el7.x86_64.rpm-report-0.txt: Adjust.
* tests/data/test-diff-pkg/spice-server-0.12.4-19.el7.x86_64-0.12.8-1.el7.x86_64-report-0.txt: Adjust.
* tests/data/test-diff-pkg/spice-server-0.12.4-19.el7.x86_64-0.12.8-1.el7.x86_64-report-1.txt: Adjust.
* tests/data/test-diff-pkg/spice-server-0.12.4-19.el7.x86_64-0.12.8-1.el7.x86_64-report-2.txt: Adjust.
* tests/data/test-read-dwarf/libtest23.so.abi: Adjust.
* tests/data/test-read-dwarf/test10-pr18818-gcc.so.abi: Adjust.
* tests/data/test-read-dwarf/test11-pr18828.so.abi: Adjust.
* tests/data/test-read-dwarf/test12-pr18844.so.abi: Adjust.
* tests/data/test-read-dwarf/test13-pr18894.so.abi: Adjust.
* tests/data/test-read-dwarf/test14-pr18893.so.abi: Adjust.
* tests/data/test-read-dwarf/test15-pr18892.so.abi: Adjust.
* tests/data/test-read-dwarf/test16-pr18904.so.abi: Adjust.
* tests/data/test-read-dwarf/test17-pr19027.so.abi: Adjust.
* tests/data/test-read-dwarf/test18-pr19037-libvtkRenderingLIC-6.1.so.abi: Adjust.
* tests/data/test-read-dwarf/test19-pr19023-libtcmalloc_and_profiler.so.abi: Adjust.
* tests/data/test-read-dwarf/test20-pr19025-libvtkParallelCore-6.1.so.abi: Adjust.
* tests/data/test-read-dwarf/test21-pr19092.so.abi: Adjust.
* tests/data/test-read-dwarf/test22-pr19097-libstdc++.so.6.0.17.so.abi: Adjust.
* tests/data/test-read-dwarf/test9-pr18818-clang.so.abi: Adjust.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
This patch speeds up the invocation of
corpus_group::get_unreferenced_function_symbols and
corpus_group::get_unreferenced_variable_symbols. It does so by
caching the result of constructing the map of of unreferenced symbol
at the time where each corpus is added to the group.
* src/abg-corpus.cc (corpus_group::unrefed_{fun, var}_symbol_map):
New data members.
(corpus_group::priv::priv): Adjust.
(corpus_group::priv::add_unref_{fun,var}_symbols): Define new
member functions.
(corpus_group::add_corpus): Update the map of unreferenced
symbols.
(corpus_group::get_unreferenced_{function,variable}_symbols)
Adjust logic.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
With this patch, kmidiff knows how to compare a serialized
corpus_group that represents a Kernel Module Interface (a .kmi file)
against either another serialized .kmi file, or against a kernel tree.
The patch extends the abixml reader to make it parse an
'abi-corpus-group' element. To do that, the patch modifies
read_corpus_from_input to make it be capable of parsing several
'abi-corpus' in a row. That modified function is then used by a new
read_corpus_group_from_input, which is itself used by the new public
entry points read_corpus_group_from_native_xml_file and
read_corpus_group_from_native_xml.
With that read_corpus_group_from_native_xml_file building block
function, the kmidiff program is modified so that it can take either
two directory trees, two .kmi files or one directory tree and one .kmi
file.
* include/abg-libxml-utils.h (advance_to_next_sibling_element):
Declare new function.
* src/abg-libxml-utils.cc (go_to_next_sibling_element_or_stay)
(advance_to_next_sibling_element): Define new functions.
* include/abg-reader.h (read_corpus_group_from_input)
(read_corpus_group_from_native_xml)
(read_corpus_group_from_native_xml_file): Declare new functions.
* src/abg-reader.cc (read_context::m_corpus_group): New data
member.
(read_context::{get_corpus_group, set_corpus_group}): Define new
member functions.
(read_translation_unit_from_input): Cleanup logic.
(read_corpus_from_input): Don't assume that the document is
starting with an 'abi-corpus' element. Support the mode where a
caller called the xmlTextReaderExpand function (and so we are
given an expanded xmlNodePtr) and the mode where we need to use
the xmlTextReader API to walk through the 'abi-corpus' element.
Also, if we are building a corpus group, do not clear what used to
be 'per-corpus' data. That data must be shared by all the corpora
of a given abi-corpus-group.
(read_corpus_group_from_input, read_corpus_group_from_native_xml)
(read_corpus_group_from_native_xml_file): Define new functions.
* include/abg-tools-utils.h (FILE_TYPE_XML_CORPUS_GROUP): New
enumerator of the file_type enum.
* src/abg-tools-utils.cc (operator<<): In the overload for
file_type, add a case for the new FILE_TYPE_XML_CORPUS_GROUP.
(guess_file_type): Dectect abi-corpus-group xml element.
* tools/abidiff.cc (adjust_diff_context_for_kmidiff): Define new
static function.
(main): Adjust to handle the new FILE_TYPE_XML_CORPUS_GROUP. That
is, compare two FILE_TYPE_XML_CORPUS_GROUP if they are present.
* tools/abilint.cc (main): Likewise.
* tools/kmidiff.cc (main): Detect that one of two .kmi files are
passed. In that case, load the .kmi file(s), build a corpus_group
of it and use it in the comparison.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
We have the kmidiff program that takes two Linux Kernel trees
containing the vmlinux binary and its modules and compare their Kernel
Module Interface, aka KMI.
We need to be able to serialize (save in a file) a representation of
that KMI. We need to load that KMI back, and compare two serialized
KMIs.
This patch implements the serialization of the KMI of a Linux Kernel
tree. It actually serializes an instance of abigail::ir::corpus_group
that is a collection of instances of abigail::ir::corpus. The KMI of
each individual binary (vmlinux or kernel module) is represented by
one abigail::ir::corpus. All the corpora share the same definitions
of types and decls, whenever that makes sense.
The patch thus factorizes the routines used to walk a Linux kernel out
of the kmidiff program. These routines are then re-used in the abidw
program to make it walk a Linux kernel tree (when the --linux-tree
option is provided), load the vmlinux and module binaries as an
instance of abigail::corpus_group and serialize it out into an output
stream.
* include/abg-tools-utils.h (check_dir)
(get_binary_paths_from_kernel_dist)
(build_corpus_group_from_kernel_dist_under): Declare new
functions. The last two functions are being moved from
tools/kmidiff.cc so that they can be re-used.
* include/abg-writer.h (write_corpus): Declare one overload that
takes a write_context parameter.
(write_corpus_group): Declare three overloads of this new function.
* src/abg-tools-utils.cc (check_dir): Define new function.
(load_generate_apply_suppressions, is_vmlinux, is_kernel_module)
(find_vmlinux_and_module_paths)
(get_binary_paths_from_kernel_dist)
(build_corpus_group_from_kernel_dist_under): Define new functions.
* src/abg-writer.cc (write_context::set_annotate): Define new
member function.
(write_corpus): Add an overload that takes a write_context. Adapt
the existing overload to make it use this new one.
(write_corpus_group): Define this new function and two additional
overloads for it.
* tools/kmidiff.cc (set_suppressions, is_vmlinux)
(is_kernel_module, find_vmlinux_and_module_paths)
(get_binary_paths_from_kernel_dist)
(build_corpus_group_from_kernel_dist_under): Remove.
(main): Adjust the call to
build_corpus_group_from_kernel_dist_under as its arguments are now
adapted since it's been factorized out into abg-tools-utils.h.
* tools/abidw.cc (options::corpus_group_for_linux): Define new
data member.
(options::options): Adjust.
(display_usage): Add help strings for the new --linux-tree option.
(load_corpus_and_write_abixml): Factorize this function out of the
main function.
(load_kernel_corpus_group_and_write_abixml): Define new function.
(main): Use the factorized load_corpus_and_write_abixml and the
new load_corpus_and_write_abixml functions.
* tests/test-read-write.cc: Adjust.
* doc/manuals/abidw.rst: Add documentation for the new
--linux-tree option.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
After the latest commits to avoid emitting empty translation units and
dupicated decls, and also to fix some indention issues in the abixml
output, this patch updates the reference outputs of the abixml
regression tests.
* tests/data/test-annotate/test13-pr18894.so.abi: Adjust.
* tests/data/test-annotate/test14-pr18893.so.abi: Adjust.
* tests/data/test-annotate/test15-pr18892.so.abi: Adjust.
* tests/data/test-annotate/test17-pr19027.so.abi: Adjust.
* tests/data/test-annotate/test19-pr19023-libtcmalloc_and_profiler.so.abi: Adjust.
* tests/data/test-read-dwarf/test13-pr18894.so.abi: Adjust.
* tests/data/test-read-dwarf/test14-pr18893.so.abi: Adjust.
* tests/data/test-read-dwarf/test15-pr18892.so.abi: Adjust.
* tests/data/test-read-dwarf/test16-pr18904.so.abi: Adjust.
* tests/data/test-read-dwarf/test17-pr19027.so.abi: Adjust.
* tests/data/test-read-dwarf/test19-pr19023-libtcmalloc_and_profiler.so.abi:
Adjust.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
It turned out a lot of empty translation unit can make it to the
abixml output. This patch reduces the number of these.
* src/abg-writer.cc (write_corpus): Do not emit a translation
unit that appears empty beforehand.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
This patch avoids emitting the same decl twice in the same ABI corpus
or corpus_group. For the purpose of this duplication detection, a
decl is designated by its pretty representation.
The patch helps to reduce the size of the abixml files drastically by
cutting some of them, especially those representing the KMI of the
linux kernel in more than half.
* src/abg-writer.cc (write_context::m_emitted_decls_map): New data
member.
(write_context::{decl_name_is_emitted, record_decl_as_emitted}):
Define new memeber functions.
(write_translation_unit): Do not emit a decl that has already been
emitted.
(write_var_decl, write_function): Record the decl as emitted.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
abidiff had an --linux-kernel-abi-whitelist option. Rename it
--kmi-whitelist.
* doc/manuals/abidiff.rst: Add documentation for the
--kmi-whitelist option.
* tools/abidiff.cc (display_usage): Emit help string for the
--kmi-whitelist option
(parse_command_line): Parse the new --kmi-whitelist option, of the
-w shortcut.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
When writting out the ABI of a Linux kernel binary we must be able to
restrict it to the set of functions and global variables which ELF
symbols have names defined in a white list.
This patch adds that support, using the --kmi-whitelist
option.
* tools/abidw.cc (options::{kabi_whitelist_paths,
kabi_whitelist_supprs}): New data members.
(display_usage): Add a help string for the new --kmi-whitelist
option.
(parse_command_line): Parse the new --kmi-whitelist option.
(maybe_check_suppression_files): Check the presence of the linux
kernel abi white list passed by the option --kmi-whitelist.
(main): Ignore loading the symbol table if the kernel abi white
list is provided.
* doc/manuals/abidw.rst: Add documentation for the new option.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
This patch makes abipkgdiff compare two kernel packages. At the
moment the comparison is done by comparing each binary from the first
package to its counterpart in the second package. No optimization is
done do represent a vmlinux binary and its modules as one single
entity. So this is different from what kmidiff does.
* include/abg-suppression.h
(variable_suppression::variable_suppression): Add default arguments
to the parameters.
* include/abg-tools-utils.h (dir_exists, dir_is_empty)
(string_begins_with, get_rpm_name, get_rpm_arch, get_deb_name)
(file_is_kernel_package, file_is_kernel_debuginfo_package):
Declare new functions.
* src/abg-tools-utils.cc (dir_exists, dir_is_empty)
(string_begins_with, get_deb_name, get_rpm_name, get_rpm_arch)
(file_is_kernel_package, file_is_kernel_debuginfo_package): Define
new functions.
(gen_suppr_spec_from_kernel_abi_whitelist): The kernel ABI
whitelist is made of ELF symbols names that ought to match
functions *and* variables that have ELF symbols with those names.
So generate variable suppression specifications as well. Not just
function suppression specifications.
* tools/abipkgdiff.cc (options::{kabi_whitelist_package,
show_symbols_not_referenced_by_debug_info, kabi_whitelist_paths,
kabi_suppressions}): New data members.
(options::options): Adjust.
(package::KIND_KABI_WHITELISTS): New enumerator in the
package::kind enum.
(package::kabi_whitelist_package_): New data member.
(package::{base_name, kabi_whitelist_package, }): New member
functions.
(display_usage): Add a help string to the new
--linux-kernel-abi-whitelist and --no-unreferenced-symbols
options.
(parse_command_line): Parse the new --no-unreferenced-symbols,
--linux-kernel-abi-whitelist and --lkaw-pkg options.
(maybe_check_suppression_files): Check the presence of kabi
whitelist files.
(set_diff_context_from_opts): Consider (not) showing symbols not
referenced by debug info.
(compare): If we are looking at linux kernel packages, take the
kernel abi whitelist into account, apply the suppressions
resulting from the kabi whitelists to the ELF read context.
(maybe_collect_kabi_whitelists)
(get_kabi_whitelists_from_arch_under_dir)
(maybe_handle_kabi_whitelist_pkg, maybe_collect_kabi_whitelists)
(get_interesting_files_under_dir): Define new functions.
(maybe_update_vector_of_package_content): Take a new
file_name_to_look_for parameter.
(create_maps_of_package_content)
(extract_package_and_map_its_content): Consider the case of the
package being a linux kernel package.
(main): Take the potential --lkaw-pkg into account.
* doc/manuals/abipkgdiff.rst: Add documentation for options
--linux-kernel-abi-whitelist, --lkaw-pkg and
--no-unreferenced-symbols.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Sometimes, it can happen that a translation unit gets read twice while
loading an abixml. This patch fixes that.
* src/abg-reader.cc (read_translation_unit): Take (in parameter) a
reference as the resulting translation unit.
(get_or_read_and_add_translation_unit): Define new static
function.
(read_context::get_scope_for_node)
(read_translation_unit_from_input): Use the new
get_or_read_and_add_translation_unit.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
* include/abg-dwarf-reader.h (set_read_context_corpus_group)
(read_and_add_corpus_to_group_from_elf, set_ignore_symbol_table)
(get_ignore_symbol_table): Declare new functions.
* abg-dwarf-reader.cc (read_context::options_type): Define new
type.
(die_dependant_container_set::clear): Define new member function.
(read_context::{bss, tesxt, rodata, data, data1}_section_): Add
new data members.
(read_context::{symbol_versionning_sections_loaded_,
symbol_versionning_sections_found_}): Likewise.
(read_context::corpus_group_): Likewise.
(read_context::{load_in_linux_kernel_mode, load_all_types,
show_stats, do_log_}): Replace these options by ..
(read_context::options_): ... this instance of the new
read_context:options_type.
(read_context::read_context): Adjust.
(read_context::{clear_alt_debug_info_data, clear_per_corpus_data,
env, get_data_section_for_variable_address, load_all_types,
load_in_linux_kernel_mode, show_stats, do_log}): Adjust.
(create_read_context): Adjust.
(read_context::~read_context): Define destructor.
(read_context::{options, bss_section, text_section,
rodata_section, data_section, data1_section, current_corpus_group,
has_corpus_group, main_corpus_from_current_group,
main_corpus_from_current_group,
current_corpus_is_main_corpus_from_current_group,
should_reuse_type_from_corpus_group}): Define new member
functions.
(read_context::get_die_qualified_type_name): Handle the name of
the current translation unit.
(read_context::load_symbol_maps): Really don't load (linux kernel
specific) symbol maps if we were told to ignore the ELF symbol
table.
(set_ignore_symbol_table, get_ignore_symbol_table)
(create_default_var_sym, create_default_fn_sym, add_symbol_to_map)
(set_read_context_corpus_group)
(read_and_add_corpus_to_group_from_elf): Define new functions.
(build_type_decl, build_typedef_type, build_enum_type)
(add_or_update_class_type)
(add_or_update_union_type): Reuse the type being built, from the
main corpus of the corpus group.
(build_qualified_type): Cleanup logic.
(build_var_decl, build_function_decl): Create a default symbol for
the variable or function if we are supposed to ignore the symbol
table of the current binary. Add that symbol to the symbol table
that is created in the read context.
(read_debug_info_into_corpus): Don't load the ELF symbol table
information if we are asked to ignore the symbol table. But set
the symbol table that we built artificially while loading
functions and variables, into the ABI corpus being built.
(read_context::maybe_adjust_var_sym_address): Adjust.
(build_ir_node_from_die): Add ir node to its logical scope. For
the C language, the scope of a type is the global scope.
(read_corpus_from_elf): Don't load ELF properties if we were asked
to avoid the ELF symbol table.
* include/abg-comparison.h (compute_diff): Declare ...
* src/abg-comparison.cc (compute_diff): ... an overload to compare
corpus_group.
* tools/kmidiff.cc: New tool.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
This patch adds support to associate a type to its source location.
The location being a string of the form "filepath:line:column". The
association is done by a per-translation unit map which associates a
location to a type.
For the moment this only associates one type to a given location. For
the general case, though, we need to associate a vector or types to a
given location. We'll add that support later.
The patch provides lookup functions, each of which looking up a
particular kind of type by its location.
* include/abg-fwd.h (get_name_of_qualified_type)
(get_name_of_reference_to_type, lookup_basic_type_per_location)
(lookup_class_type_per_location, lookup_union_type_per_location)
(lookup_enum_type_per_location, lookup_typedef_type)
(lookup_typedef_type_per_location, lookup_pointer_type)
(lookup_reference_type, lookup_type_per_location)
(lookup_type_through_translation_units)
(lookup_type_from_translation_unit, odr_is_relevant): Declare new
functions or new function overloads.
* include/abg-ir.h (location::expand): Declare new member
function.
(type_maps::empty): Likewise.
(operator|=): Declare an overload for qualified_type_def::CV.
(get_string_representation_of_cv_quals)
(get_name_of_qualified_type, lookup_qualified_type): Declare new functions.
* src/abg-ir.cc (location::expand): Define new member function.
(type_maps::empty): Likewise.
(odr_is_relevant): Likewise.
(get_string_representation_of_cv_quals)
(get_name_of_reference_to_type, get_name_of_qualified_type)
(lookup_union_type_per_location): Define new functions or overloads.
(lookup_basic_type, lookup_enum_type, lookup_typedef_type)
(lookup_qualified_type, lookup_pointer_type)
(lookup_reference_type, lookup_type_from_translation_unit)
(lookup_basic_type_per_location, lookup_basic_type_per_location)
(lookup_class_type_per_location, lookup_class_type_per_location)
(lookup_enum_type_per_location, lookup_enum_type_per_location)
(lookup_typedef_type_per_location)
(lookup_typedef_type_per_location, lookup_type_per_location):
Define new overloads.
(maybe_update_types_lookup_map)
(maybe_update_types_lookup_map<class_decl>)
(maybe_update_types_lookup_map<function_type>): Add a new
use_type_name_as_key parameter. If it's false, then associates
the type to its location rather than to its name.
(maybe_update_types_lookup_map): In the overloads for type_decl,
class_decl, union_decl, enum_type, typedef_decl, array_type_def,
record the type in the lookup map per location, in addition to the
per-name recording.
(qualified_type_def::build_name): Use the new
get_name_of_qualified_type.
(qualified_type_def::get_cv_quals_string_prefix): Use the new
get_string_representation_of_cv_quals.
(operator|=): Define a new overload for qualified_type_def::CV.
(pointer_type_def::get_qualified_name): Use the new
get_name_of_pointer_to_type.
(reference_type_def::get_qualified_name): Use the new
get_name_of_reference_to_type.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
To support the upcomping analysis of the Linux kernel and its modules,
we need a way to represent a union of corpora. The first corpus
loaded would be the one representing the vmlinux binary. Subsequent
corpora loaded would be those representing the modules.
This patch provides the new abigail::ir::corpus_group type that
represents such a corpus group.
* include/abg-corpus.h (corpus::{find_translation_unit,
get_type_per_loc_map}): Declare new member functions.
(corpus::{get_architecture_name, is_empty}): Make these member functions
const.
(corpus::{get_sorted_fun_symbols, get_functions, get_variables,
get_unreferenced_function_symbols,
get_unreferenced_variable_symbols}): Make these member functions
virtual.
(class corpus_group): Declare a new type.
* include/abg-fwd.h (corpus_sptr, corpus_group_sptr)
(string_tu_map_type, istring_var_decl_ptr_map_type)
(istring_function_decl_ptr_map_type): Define new typedefs.
* src/abg-corpus-priv.h (corpus_priv::{path_tu_map,
type_per_loc_map_}): Add new data members.
* src/abg-corpus.cc (corpus_add): Complete the function comment.
Assert that at most one translation unit of a given path can be
added to the corpus.
(corpus::{find_translation_unit, get_type_per_loc_map}): Define
new member functions.
(corpus::{get_architecture_name}): Make this member function
const.
(struct corpus_group::priv): Define new type.
(corpus_group::{corpus_group, ~corpus_group, add_corpus,
get_corpora, is_empty, get_functions, get_variables,
get_var_symbol_map, get_fun_symbol_map, get_sorted_fun_symbols,
get_sorted_var_symbols, get_unreferenced_function_symbols,
get_unreferenced_variable_symbols}): Define member functions of
the new corpus_group type.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
The command "fedabipkgdiff --self-compare --from fc25 grub2" returns
with an error because no debug info can be found for the file
kernel.img. That file doesn't have any debug info shipped for it.
And we shouldn't try to compare it anyway.
This patch updates the default suppression file shipped by libabigail
to make it avoid compare kernel.img files.
* default.abignore: Do not compare kernel.img files.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
The build libxcb-1.12-3.fc26 has been tagged for Fedora 27 (i.e,
fc27).
When we ask fedabipkgdiff to get the builds of libxcb for Fedora 27,
it looks for builds which release string ends with 'fc27'. It thus
fails to pick libxcb-1.12-3.fc26.
This patch changes this behaviour by making
Brew.get_package_latest_build to first try to get builds which
release string match exactly the distro string we are looking at.
But then if no build was found, the member function then tries to get
builds for which the distro part of the release string is "less than"
(in a lexicographic way) the distro string we are looking at.
I haven't added any regression test specifically for this because we
are planning to use the libabigail-selfcheck external tool to perform
regression testing on tons of Fedora packages:
https://pagure.io/libabigail-selfcheck. Fixing this issue is a
pre-requisite for putting that regression test infrastructure in
place.
* tools/fedabipkgdiff (get_distro_from_string): Define new function.
(Brew.get_package_latest_build): Also consider builds which distro
property is less than the expected distro string that we were
given.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
In fedabipkgdiff tool, some packages might have only
noarch sub-packages. In these cases, no package is
available for running abipkgdiff. This leads to
return_codes list being empty.
* tools/fedabipkgdiff (run_abipkgdiff()): Check if
return_codes list is empty
Signed-off-by: Sinny Kumari <sinny@redhat.com>
When abipkgdiff decides that two packages have no content to compare
it forgets to remove the temporary directories that were created.
* tools/abipkgdiff.cc (maybe_erase_temp_dirs): Define new static
function.
(compare): Call the new maybe_erase_temp_dirs on all return
points.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
* include/abg-ini.h (config::section::priv): Make this be a class,
not a struct.
* src/abg-dwarf-reader.cc (build_translation_unit_and_add_to_ir)
(build_ir_node_from_die): Add parenthesis around assignment
expressions inside conditional expression.
* src/abg-suppression.cc (read_function_suppression): Likewise.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
* include/abg-comp-filter.h (class filter_base): Declare this as a
struct.
* include/abg-comparison.h (class filtering::filter_base):
Likewise.
(struct diff_traversable_base): Declare this as a class.
* include/abg-ir.h (function_decl::parameter): Declare this before
using it.
* src/abg-corpus.cc
(corpus::priv::build_unreferenced_symbols_tables): Add missing
parenthesis around assignment expressions inside conditional
expressions.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
fedabipkgdiff tool can communicate with Fedora koji and has
capability to download and perform ABI comparison between
specified NVRs.
With addition of --self-compare option, it will be possible
to perform ABI comparison on same package. One of the important
usecase of this option is to run automated ABI checks on
packages with known expected result i.e. no ABI change. This usecase
will be useful to ensure that libabigail functionality doesn't
break with new commits made.
This option can be invoked as:
fedabipkgdiff -a --self-compare -fc24 package_name
* bash-completion/fedabipkgdiff: Add new option --self-compare
* tests/data/Makefile.am: Add new test file
* tests/data/test-fedabipkgdiff/test7-self-compare-from-fc23-dbus-glib-report-0.txt:
New reference output for testing ABI comparison on same package
* tests/runtestfedabipkgdiff.py.in (FEDABIPKGDIFF_TEST_SPECS):
Add test case for --self-compare
* tools/fedabipkgdiff (build_commandline_args_parser()): Add
new option --self-compare
(generate_comparison_halves()): Find second comparision half in same
package list while doing self-compare
(self_compare_rpms_from_distro()): New function to perform ABI
comparision on same pacakge
(main()): Add if condition when --self-compare option is enabled
The C language doesn't support namespaces. So, in that language, the
context of a type is always the global scope. Hence, the context of a
type DIE is always the global context associated to the current
translation unit.
Thus, when we are analyzing a C binary, we can do away with building
the DIE -> DIE parent map that we later use to get the parent of a
type DIE when we need to determine the context of a given type DIE.
This can save significant time and space.
This patch implements this optimization.
* include/abg-ir.h (global_scope_sptr): Make this be a share_ptr
of scope_decl, not of global_scope.
(translation_unit::get_global_scope): Return a reference to
scope_decl_sptr.
* src/abg-ir.cc (translation_unit::get_global_scope): Return a
scope_decl not a global_scope.
* src/abg-dwarf-reader.cc (read_context::nil_scope_): Add new data
member.
(read_context::{global_scope, nil_scope}): Define new member functions.
(read_context::build_die_parent_maps): Do not build the map if we
are looking at a C (or asm) translation unit.
(get_scope_die, get_scope_for_die): If we are looking at a C
translation unit then do return the global scope.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
While emitting an abixml output for a corpus_group representing the
KMI of a Linux kernel tree, profiling shows that during comparison of
class types, calling class_decl::get_definition_of_declaration is a
hotspot. Especially, the fact the function returns a shared pointer
that has to be "handled" shows up in the profile.
This patch introduces a
class_decl::get_naked_definition_of_declaration that returns a
pointer. Not a shared pointer.
The patch then uses that get_naked_definition_of_declaration function
in the comparison code for class_decl.
The patch also uses get_naked_canonical_type instead of
get_canonical_type when comparing a bunch of other types.
This makes things a little bit faster when compiled without
optimization between 2% and 5%.
* include/abg-ir.h
(class_or_union::get_naked_definition_of_declaration): Declare a
new member function.
(class_decl::get_naked_definition_of_declaration): Likewise.
* src/abg-ir.cc ({type_decl, qualified_type_def,
array_type_def, enum_type_decl}::operator==): Use the
get_naked_canonical_type and get_naked.
(class_or_union::priv::naked_definition_of_declaration_): Define
new data member.
(class_or_union::priv::priv): Adjust to initialize the new data
member.
(class_or_union::get_naked_definition_of_declaration): Define new
member function.
({class_or_union, class_decl}::operator==): Use the new
get_naked_definition_of_declaration instead of
get_definition_of_declaration.
(equals): In the overload for class_or_union, do the same.
(class_decl::get_naked_definition_of_declaration): Define new
member function.
(hash_type_or_decl): Likewise.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Profiling shows that while writting out the abixml for a big
corpus_group, comparing decl-only classes is a hot spot.
This patch avoids calling the function
class_or_union::priv_->comparison_started (which appears to be
culprit) in that case.
This makes writting out the abixml of the corpus_group of the Linux
kernel be ~20% faster.
* src/abg-ir.cc (equals): In the overload for class_decl, if we
are looking at a decl-only class, then directly call the equals
function for class_or_union. That one knows how to perform the
comparison without calling the
class_or_union::priv_->comparison_started function, in that case.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
This updates the description of what abipkgdiff in termes of sequence
of actions.
* tools/abipkgdiff.cc: Update the description of the sequence of
actions performed.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
* src/abg-writer.cc (class write_context): Fix indentation.
(write_location, write_visibility, write_binding)
(write_array_size_and_alignment, write_size_and_alignment): Fix
these declarations to use the *_sptr typedefs rather than the
explicit shared_ptr<*> types.
(write_translation_unit): Fix comment.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Suppose we have two types struct A and struct B. Suppose the two
types have members of integer type. Suppose there are members of
integer type in struct A that are modified to become members of char
type. Suppose, at last, that we also have members of integer type in
struct B that are modified to become members of char type.
In that case, we want to see all the changes of members which types
got changed from integer type to char type. None of these changes
should be considered redundant.
Today, unfortunately, only the first change is reported by default.
The subsequent changes are considered to be redundant.
This patch fixes that by considering that changes that modifies the type of a
decl from a basic type into another are never considered redundant.
* include/abg-comparison.h (is_diff_of_basic_type)
(has_basic_type_change_only): Declare these functions ...
* src/abg-comparison.cc (is_diff_of_basic_type)
(has_basic_type_change_only): ... and define them.
(redundancy_marking_visitor::visit_begin): Use the new
has_basic_type_change_only.
* tests/data/test-diff-filter/libtest37-v0.so: New binary test input.
* tests/data/test-diff-filter/libtest37-v1.so: Likewise.
* tests/data/test-diff-filter/test37-report-0.txt: New test
reference output.
* tests/data/test-diff-filter/test37-v0.cc: Source code of the new
binary test input.
* tests/data/test-diff-filter/test37-v1.cc: Likewise.
* tests/data/Makefile.am: Update to add the new test material to
the source distribution.
* tests/test-diff-filter.cc (in_out_spec): Add the new test input
to this test harness.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
When {function, var}_decl::set_symbol is invoked, the cache of of the
ID of the decl must be invalidated because that function changes the
ID of the decl.
This patch does just that.
* src/abg-ir.cc ({function, var}_decl::set_symbol): Invalidate the
ID cache.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Now that there is a is_type predicate that takes a a type_or_decl_base
type, the overloads that take a decl_base or a type_base are useless
and can even lead to overload resolution issues. This patch removes
those useless overloads.
* include/abg-fwd.h (is_type): Remove the overloads that take
decl_base and type_base types.
* src/abg-ir.cc (is_type): Likewise.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
Const references are a useless redundancy because a reference is
always const.
Thus build_qualified_type can return a non-qualified type when it
figures out the qualifier is useless.
This patch adjusts build_qualified_type to recognize that. It makes
it return a type_base_sptr rather than a qualified_type_def_sptr.
In build_ir_node_from_die, in the case where we handle qualified
types, the call to build_qualified_type is adjusted accordingly.
* src/abg-dwarf-reader.cc (build_qualified_type): Return a
type_base_sptr.
(build_ir_node_from_die): Adjust the call to build_qualified_type.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
In some case, some array subranges would all have the same lower
bound.
Fixed thus.
* src/abg-dwarf-reader.cc (build_subranges_from_array_type_die):
Consider the 'lower_bound' parameter as the default lower bound
for each sub-ranges.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
I noticed a buffer overrun in the equals overload for arrays. This
patch fixes that.
* src/abg-ir.cc (equals): In the overload for arrays, check for
the end of the subranges of the two arrays, not just for the first
one.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
It can happen that queue::priv_::do_bring_workers_down stays forever
waiting for a task to finish (via pthread_join). The it waits for is
itself blocked in worker::wait_to_execute_a_task, in pthread_cond_wait.
It seems to me that this is because we forget to lock the
queue::priv::queue_cond_mutex before inspecting and updating the
variables on which the wait on the condition depend.
This patch fixes that.
The patch also moves tests/test-read-write.cc over to using the work queue
to increase test coverage for the work queue interface.
* src/abg-workers.cc (queue::priv::tasks_todo_mutex): Make this
data member mutable.
(more_tasks_to_execute):
(queue::priv::do_bring_workers_down): Update the
queue::priv::bring_workers_down only in the critical section
defined by queue::priv::queue_cond_mutex.
(worker::wait_to_execute_a_task): Testing for
queue::priv::bring_workers_down is done in the critical section
defined by queue::priv::queue_cond_mutex. The loop over waiting
ont the condition is also in the critical section, as it ought to
be.
* tests/test-read-write.cc (struct test_task): New type.
(main): Express in terms of the new test_task type.
Signed-off-by: Dodji Seketeli <dodji@redhat.com>
