Fix attaching to re-prelinked executables on ppc64.
* solib-svr4.c (svr4_exec_displacement): New variable plt2_asect,
initialize it, try to adjust FILESZ field by it.
Fix missing _start PIE relocation on ppc64 due to stop on dl_main.
* break-interp.exp (reach): Move the core body ...
(reach_1): ... here. Use stop-on-solib-events for _dl_debug_state.
(test_ld): Provide always real argument, even to the linker. Replace
dl_main by _dl_debug_state.
(thumb_instruction_changes_pc): Likewise.
(thumb2_instruction_changes_pc): Likewise.
(thumb_analyze_prologue): Handle 32-bit Thumb instructions during
prologue parsing. Improved support for optimized code.
(thumb_scan_prologue): Do not reply on line-number information,
use same heuristics as arm_scan_prologue insead.
(skip_prologue_function): Accept functions
"__tls_get_addr" and "__aeabi_read_tp".
symbol_table and index_table_slots to symbol_table_slots.
All uses updated.
(create_symbol_hash_table): Renamed from create_index_table.
All callers updated.
(add_indices_to_cpool): Rename arg index_table to symbol_hash_table.
(write_hash_table): Rename local index_table to symbol_hash_table.
(addrmap_foreach): Declare.
* addrmap.c (struct addrmap_funcs): New member foreach.
(addrmap_foreach): New function.
(addrmap_fixed_foreach): New function.
(addrmap_fixed_funcs): Update.
(struct mutable_foreach_data): New struct.
(addrmap_mutable_foreach_worker): New function.
(addrmap_mutable_foreach): New function.
(addrmap_mutable_funcs): Update.
Several changes to dict_hash from Adacore's local version were inadvertantly
left out of the last change to that function. These changes conservatively
cause dict_hash to revert to msymbol_hash_iw in more cases where the symbol
being hashed is known not to be Ada-encoded. This avoids a few cases where
the function hashes large groups of names with similar prefixes to the same
value.
Changelog:
gdb/
* dictionary.c (dict_hash): Revert to msymbol_hash_iw in
more cases.
This is a clean-up of some of our symbol-lookup machinery to pull some
kludgy Ada-specific definitions out of psymtab.c. In place of
map_ada_symtabs and ada_lookup_partial_symbol, we have a method
map_matching_symbols, which searches through all symbol tables and
partial symbol tables looking for a symbol that matches according to
a matching function that is passed as a parameter. This requires some
care, because partial symbol tables speed up searches by binary search,
while full symbol tables use hashing. To call map_matching_symbols, therefore,
you may need to supply both a matching function that is compatible with the
dictionary hash function and an ordering relation that is compatible with
strcmp_iw, which is used to order partial symbol tables.
Having added this general routine to psymtab.c, we use it in ada-lang.c
to rework add_non_local_symbols (now renamed add_nonlocal_symbols).
Changelog:
gdb/
* ada-lang.c (full_match): Declare.
(ada_match_name): Rename to match_name (we should avoid prefixing static
symbols with "ada_").
(match_name): New name for ada_match_name.
(struct ada_psym_data): Remove and replace with...
(struct match_data): User data for map_matching_symbols.
(ada_add_psyms): Remove.
(aux_add_nonlocal_symbols): New function, used as callback for
map_matching_symbols.
(compare_names): Ordering function adopted from strcmp_iw for Ada-encoded
symbols.
(ada_add_non_local_symbols): Rename to add_nonlocal_symbols.
(add_nonlocal_symbols): Renamed from ada_add_non_local_symbols.
Rework to use map_matching_symbols instead of map_ada_symtabs.
(ada_lookup_symbol_list): Use add_nonlocal_symbols.
* psymtab.c: Include dependency on dictionary.h.
(match_partial_symbol): New function.
(ada_lookup_partial_symbol): Remove.
(map_block): New function, auxiliary to map_matching_symbols_psymtab.
(map_matching_symbols_psymtab): New function.
(psym_functions): Replace map_ada_symtabs with map_matching_symbols_psymtab.
* symfile.h: Replace map_ada_symtabs definition with map_matching_symbols.
This patch allows Ada to speed up symbol lookup by using the facilities
in dictionary.[ch] for hashed lookups. First, we generalize dictionary
search to allow clients to specify any matching function compatible with
the hashing function. Next, we modify the hashing algorithm so that symbols
that wild-match a name hash to the same value. Finally, we modify Ada
symbol lookup to use these facilities.
Because this patch touches on a hashing algorithm used by other
languages, I took the precaution of doing a speed test on a list of
about 12000 identifiers (repeatedly inserting all of them into a table
and then doing a lookup on a million names at random, thus testing the
speed of the hashing algorithm and how well it distributed names).
There was actually a slight speedup, probably as a result of open-
coding some of the tests in msymbol_hash_iw. By design, the revised
hashing algorithm produces the same results as the original on most
"normal" C identifiers.
We considered augmenting the dictionary interface still further by allowing
different hashing algorithms for different dictionaries, based on the
(supposed) language of the symbols in that dictionary. While this produced
better isolation of the changes to Ada programs, the additional flexibility
also complicated the dictionary interface. I'd prefer to keep things
simple for now.
Tested w/o regressions on Linux i686.
ChangeLog:
gdb/
* ada-lang.c (ada_match_name): Use new API for wild_match.
(wild_match): Change API to be consistent with that of strcmp_iw;
return 0 for a match, and switch operand order.
(full_match): New function.
(ada_add_block_symbols): Use dict_iter_match_{first,next} for
matching to allow use of hashing.
* dictionary.c (struct dict_vector): Generalize iter_name_first,
iter_name_next ot iter_match_first, iter_match_next.
(iter_name_first_hashed): Replace with iter_match_first_hashed.
(iter_name_next_hashed): Replace with iter_match_next_hashed.
(iter_name_first_linear): Replace with iter_match_first_linear.
(iter_name_next_linear): Replace with iter_match_next_linear.
(dict_iter_name_first): Re-implement to use dict_iter_match_first.
(dict_iter_name_next): Re-implement to use dict_iter_match_next.
(dict_iter_match_first): New function.
(dict_iter_match_next): New function.
(dict_hash): New function.
* dictionary.h (dict_iter_match_first, dict_iter_match_next): Declare.
* psymtab.c (ada_lookup_partial_symbol): Use new wild_match API.
* Makefile.in (XML_SYSCALL_DIR, XML_SYSCALL_FILES): Moved to
data-directory/Makefile.in.
(SUBDIRS): Add data-directory.
(all): Remove xml-syscall-copy dependency.
(xml-syscall-copy): Moved to data-directory/Makefile.in as
stamp-syscalls.
(xml-syscall-install): Moved to data-directory/Makefile.in as
install-syscalls.
(install-only): Remove xml-syscall-install dependency.
(all-data-directory): New rule.
(data-directory/Makefile): New rule.
* configure.ac (AC_OUTPUT): Add data-directory/Makefile.
* configure: Regenerate.
* data-directory/Makefile.in: New file.
* python/lib/gdb/__init__.py: New file.
testsuite/
* gdb.base/catch-syscall.exp (do_syscall_tests): Update location
of syscall staging area.
When one of the parameter values in a subprogram calls is an array
whose value does not come from inferior memory, the debugger first
copies the array value in inferior memory. Up to now, the memory
used to hold that copy was taken from the stack (just below the SP),
but this is causing problems on SPARC v9.
So the immediate fix is to follow what C does with arrays and strings,
which is allocate memory on the heap.
gdb/ChangeLog:
* ada-lang.c: #include "value.h".
(ensure_lval): Delete advance declaration. Remove gdbarch and sp
arguments. Implement using value_allocate_space_in_inferior
instead of allocating memory from the stack.
(make_array_descriptor): Remove gdbarch and sp parameters. Update
calls to ensure_lval.
(ada_convert_actual): Remove gdbarch and sp parameters. Update
calls to make_array_descriptor and ensure_lval.
* ada-lang.h (ada_convert_actual): Update declaration.
* infcall.c (value_arg_coerce): Update call to ada_convert_actual.
When a type is marked with pragma Unchecked_Variants, ptype did not print
variants having a single component, since the compiler produces incorrect
debugging output for such cases. With this patch, we special-case these
components so that they print.
Changelog:
* gdb/ada-typeprint.c (print_selected_record_field_types): New function,
incorporating and generalizing print_record_field_types.
(print_record_field_types): Change return value and update comment.
Re-implement using print_selected_record_field_types.
(print_choices): Print "=>" here.
Handle case of unencoded variant branch.
(print_variant_clauses): Reformat comment.
Special-case unencoded variant branch.
This new version of wild_match is comparable in speed to strcmp_iw, and has
the same signature and same return value for equal names.
gdb/ChangeLog:
* ada-lang.c (wild_match): Reimplement.
Change API to eliminate unused length argument, reverse arguments and
make 0 the 'true' return value.
(advance_wild_match): New auxiliary function for wild_match to improve
readability.
(ada_match_name, ada_add_block_symbols): Use new API for wild_match.
* psymtab.c (ada_lookup_partial_symbol, map_ada_symtabs): Use new
API for wild_match.
* symfile.h (map_ada_symtabs): Modify declaration to use new API for
wild_match.
* dwarf2read.c (dw2_map_ada_symtabs): Ditto.
Jan Kratochvil
Ulrich Weigand
Pierre Muller
Matthew Gretton-Dann
gdbadmin
Doug Evans
Daniel Jacobowitz
Sami Wagiaalla
Ken Werner
Paul N. Hilfinger
Joel Brobecker
Maciej W. Rozycki