type Char_Table is array (Character range Character'First .. Character'Last)
of Natural;
Trying to print the type description of this type currently yields:
(gdb) ptype char_table
type = array ('["00"]' .. '["ff"]') of natural
Although technically correct, it seemed more useful to print the array
range as:
(gdb) ptype char_table
type = array (character) of natural
This patch implements this suggestion.
gdb/ChangeLog:
* ada-typeprint (type_is_full_subrange_of_target_type):
New function.
(print_range): Add parameter bounds_prefered_p. If not set,
try printing range types using the name of their base type.
(print_range_type): Add parameter bounds_prefered_p.
Use it in call to print_range.
(print_array_type, ada_print_type): Update calls to print_range
and print_range_type.
gdb/testsuite/ChangeLog:
* gdb.ada/array_char_idx: New testcase.
These declarations are unncessary, and make it extra work when trying
to change the profile of one of these functions. This patch just
removes them.
Note that one of them (print_dynamic_range_bound), was improperly
indented.
gdb/ChangeLog:
* ada-typeprint.c (print_array_type, print_choices, print_range)
(print_range_bound, print_dynamic_range_bound, print_range_type):
Remove declaration.
Consider the following declarations:
type Range_Type is (One, Two, Three);
type Array_Type is array (Range_Type range One .. Two) of Integer;
A : Array_Type := (1, 2);
Trying to print A can yield:
(gdb) print a
$1 = (one => 1, 2)
The bound of the first element should not have been printed, since
"one" is the first enumerate of type Range_Type. Similarly, with
the following declarations:
type Array2_Type is array (Range_Type range Two .. Three) of Integer;
A2 : Array2_Type := (2, 3);
GDB is failing to print the bound of the first element of "A2":
(gdb) print a2
$2 = (2, 3)
This is because the index type for both types Array_Type and Array2_Type
are subranges (by DWARF definition for arrays), of an anonymous subrange
type. When deciding whether to print the bound of the first element,
we handle subranges, but only up to one level. This patch enhanced
the code to handle any number of subrange levels.
gdb/ChangeLog:
* ada-valprint.c (print_optional_low_bound): Get index_type's
target type for as long as it is a TYPE_CODE_RANGE.
No testcase with this patch, but this will be tested via the testcase
of another patch, which uses the DWARF assembler to generate debugging
info for an array indexed by an enum.
On x86-solaris, the gcore command sometimes triggers the following
internal error:
(gdb) gcore
/[...]/procfs.c:5523: internal-error: procfs_make_note_section: Assertion `thread_args.note_data != note_data' failed.
The problem is extremely elusive, for reasons that will become clearer
as I explain what is going on.
The program used to produce this issue was really simple:
| void break_me (void) { }
|
| int
| main (void)
| {
| break_me ();
| return 0;
| }
The procfs_make_note_section builds a buffer incrementally with
the contents of the core's notes section. The interesting bits are:
char *note_data = NULL;
[...]
note_data = (char *) elfcore_write_prpsinfo (obfd,
note_data,
note_size,
fname,
psargs);
This is the first call to bfd's elfcore which initializes note_data.
After that, we have a few more calls, which keep updating notes_data
and note_size, but our interest lies in the following part of
the function:
thread_args.note_data = note_data;
[...]
proc_iterate_over_threads (pi, procfs_corefile_thread_callback,
&thread_args);
/* There should be always at least one thread. */
gdb_assert (thread_args.note_data != note_data);
The comment implies that the assert is to verify that our loop
iterated over at least one thread. The check is relying on the
fact that the notes_data returned by the elfcore module changes
at each iteration, via (in procfs_corefile_thread_callback):
args->note_data = procfs_do_thread_registers (args->obfd, ptid,
args->note_data,
args->note_size,
args->stop_signal);
(which calls elfcore_write_lwpstatus).
But, while it happens most of the time, thanks to a call to realloc
in elfcore_write_note (the function that actually appends the data
at the end of the notes buffer),...
buf = (char *) realloc (buf, *bufsiz + newspace);
... this is by no means guarantied. In fact, under the right
circumstances, the buffer was grown twice without changing
addresses. Unfortunately, the circumstances are very sensitive,
thus making this bug very elusive.
This patch fixes the problem by simply removing the assert.
This means we're losing the assertion that there is at least one
thread, but I think that's OK. If we still want to keep the
assertion, we have the option of either checking the buffer
size, or else adding a boolean flag in the context structure
that we'd set to true as soon as we have a thread.
gdb/ChangeLog:
* procfs.c (procfs_make_note_section): Remove assertion and
associated comment.
For each object, if it has a nonempty .data or .bss section,
emit a symbol that could cause the startup code to selectively
link in the code to initialize those sections.
* config/tc-msp430.c (msp430_section): Always flag data sections,
regardless of -md.
(msp430_frob_section): New. Make sure all sections are noticed if
they have content.
(msp430_lcomm): New. Flag bss if .lcomm is seen.
(msp430_comm): New. Likewise.
(md_pseudo_table): Add them.
* config/tc-msp430.h (msp430_frob_section): Declare.
(tc_frob_section): Define.
Functions remote_read_bytes and get_core_siginfo are the callees of
target to_xfer_partial interface, so argument 'len' should be changed
to type ULONGEST.
gdb:
2014-01-24 Yao Qi <yao@codesourcery.com>
* remote.c (remote_read_bytes): Change type of len to ULONGEST.
* corelow.c (get_core_siginfo): Likewise.
Hi,
This patch changes the type of 'len' from ssize_t to ULONGEST.
At the beginning Siddhesh Poyarekar proposed this patch
[PATCH] Memory reads and writes should have size_t length
https://sourceware.org/ml/gdb-patches/2012-05/msg01073.html
to change type of 'len' to size_t. However, after Jan's review, we
decide to change it to ssize_t, because callers of these functions
may pass signed type to them.
AFAICS, the target layer is a boundary. In one side, we pass size_t
or ssize_t to target related APIs, and in the other side, the
implementation side, we used LONGEST (ULONGEST in latest code) because
of to_xfer_partial.
Since remote_write_bytes_aux and remote_write_bytes belong to the
implementation of remote target, we should use ULONGEST for len, IMO.
Regression tested on x86_64-linux. Is it OK?
gdb:
2014-01-24 Yao Qi <yao@codesourcery.com>
* remote.c (remote_write_bytes_aux): Change type of 'len' to
ULONGEST. Don't check 'len' is negative.
(remote_write_bytes): Change type of 'len' to ULONGEST.
Ensures TLS orphans are placed adjacent to existing TLS sections,
and fixes places where the output_section_statement flags (which might
not be set) were tested when bfd_section flags were available.
* ldlang.c (lang_output_section_find_by_flags): Be careful to
test look->bfd_section->flags if available rather than
look->flags. Separate SEC_THREAD_LOCAL handling from
SEC_READONLY loop, and rewrite.
ELFv2 needs fewer relocs to annotate plt call stubs. I correctly
allocated a smaller buffer and wrote the proper relocs, but stupidly
bumped the reloc count as for ELFv1.
* elf64-ppc.c (ppc_build_one_stub): Correct reloc count passed
to get_relocs for ELFv2.
New MSP430 MCU parts are being created by TI all the time and the
list is basically always out of date. Instead any name will be
accepted by the -mmcu= command line option. ISA selection is now
based upon the -mcpu= command line option, just as is done for GCC.
gas/ChangeLog
* config/tc-msp430.c (show_mcu_list): Delete.
(md_parse_option): Accept any MCU name. Accept several more
variants for the -mcpu option.
(md_show_usage): Do not call show_mcu_list.
Bad linker script may lead to TLS sections separated by non-TLS sections
in output. This patch changes linker assert to a linker error to
provide better linker diagnosis.
PR ld/16498
* elf.c (_bfd_elf_map_sections_to_segments): Issue a linker error
if TLS sections are not adjacent.
This fixes a bug in FrameDecorator.py.
FrameVars seems to assume that Frame.block can return None if there is
no block. However, it actually throws an exception.
I saw this bug while developing a frame filter, but unfortunately I
don't know how to reproduce it. It seems to me that the SAL tests in
_is_limited_frame should exclude the bad cases; and in my attempts to
write a test they do.
Nevertheless I think the fix is reasonably obvious and ought to go in.
2014-01-23 Tom Tromey <tromey@redhat.com>
PR python/16485:
* python/lib/gdb/FrameDecorator.py: (FrameVars.fetch_frame_args):
Handle exception from frame.block.
(FrameVars.fetch_frame_locals): Likewise.
This fixes PR python/16487.
The bug here is that the function-name-handling code in py_print_frame
had a small logic error (really a misplaced closing brace). This
error could lead to a Py_DECREF(NULL), which crashes.
This patch fixes the bug in the obvious way.
Built and regtested on x86-64 Fedora 18. New test case included.
2014-01-23 Tom Tromey <tromey@redhat.com>
PR python/16487:
* python/py-framefilter.c (py_print_frame): Don't call Py_DECREF
on a NULL pointer. Move "goto error" to correct place.
2014-01-23 Tom Tromey <tromey@redhat.com>
PR python/16487:
* gdb.python/py-framefilter.exp: Add test using "Error" filter.
* gdb.python/py-framefilter.py (ErrorInName, ErrorFilter): New
classes.
apply_frame_filter calls ensure_python_env before computing the
gdbarch to use. This means that python_gdbarch can be NULL while in
Python code, and if a frame filter depends on this somehow (easy to
do), gdb will crash.
The fix is to compute the gdbarch first.
Built and regtested on x86-64 Fedora 18.
New test case included.
2014-01-23 Tom Tromey <tromey@redhat.com>
PR python/16491:
* python/py-framefilter.c (apply_frame_filter): Call
ensure_python_env after computing gdbarch.
2014-01-23 Tom Tromey <tromey@redhat.com>
PR python/16491:
* gdb.python/py-framefilter.py (Reverse_Function.function): Read a
string from an inferior frame.
* gdb.python/py-framefilter-mi.exp: Update.
This patch changes the argument type to gdb_byte * in order to align
with the to_xfer_partial interface.
gdb:
2014-01-23 Yao Qi <yao@codesourcery.com>
* target.c (raw_memory_xfer_partial): Change argument type
from void * to gdb_byte *.
(memory_xfer_partial_1, memory_xfer_partial): Likewise.
* config/tc-msp430.c (msp430_refsym): New: ".refsym <symbol>"
* doc/c-msp430.texi (MSP430 Directives): Document it.
The purpose of this patch is to provide a way for one object file
to require the inclusion of another object, without having to
allocate space for a .word address reference.
Since regzmm can't be used in AVX2 gather instructions, there is no need
to check regzmm in AVX2 gather assert.
2014-01-22 Michael Zolotukhin <michael.v.zolotukhin@gmail.com>
* config/tc-i386.c (check_VecOperands): Remove regzmm from AVX2
gather assert.
* syscalls/s390x-linux.xml: New file.
* syscalls/s390-linux.xml: New file.
* s390-linux-tdep.c (XML_SYSCALL_FILENAME_S390): New macro.
(XML_SYSCALL_FILENAME_S390X): Likewise.
(op_svc): New enum value for SVC opcode.
(s390_sigtramp_frame_sniffer): Replace literal by 'op_svc'.
(s390_linux_get_syscall_number): New function.
(s390_gdbarch_init): Register '*get_syscall_number' and the
syscall xml file name.
* data-directory/Makefile.in (SYSCALLS_FILES): Add
"s390-linux.xml" and "s390x-linux.xml".
* NEWS: Announce new feature.
gdb/testsuite/ChangeLog:
* gdb.base/catch-syscall.exp: Activate test on s390*-linux.
The trace-specific test case 'entry-values' concludes fairly late in
the process that this platform doesn't support trace. Before that,
there are some platform specifics that don't work on s390x. The fix
addresses two aspects:
(1) Removal of an excess space character in the regex for the
disassembly. This is needed when there is a function alignment
gap, because then the hex address is immediately followed by a
colon, like in the first 'nopr' line below:
(gdb) disassemble foo+50,+10
Dump of assembler code from 0x32 to 0x3c:
0x0000000000000032 <foo+50>: br %r4
0x0000000000000034: nopr %r7
0x0000000000000036: nopr %r7
0x0000000000000038 <bar+0>: stmg %r11,%r15,88(%r15)
End of assembler dump.
(2) Handling for the s390-specific call instruction.
gdb/testsuite/ChangeLog:
* gdb.trace/entry-values.exp: Remove excess space character from
regex patterns. Handle s390 call instruction.
On ppc64-linux a function symbol does not point to code, but to the
function descriptor. Thus the previous change for this test case
broke it:
https://sourceware.org/ml/gdb-patches/2014-01/msg00275.html
This patch reverts to the original method, re-introducing '_start'
symbols. In addition, it adds sufficient alignment before the label,
such that the label never points into an alignment gap.
gdb/testsuite/ChangeLog:
* gdb.dwarf2/dw2-dir-file-name.c (FUNC): Insert alignment and
define "*_start" label. Make "name" static.
* gdb.dwarf2/dw2-dir-file-name.exp: Replace references to
${name} by references to ${name}_start.
When upstream gcc is given a command line with the "-g" option after
"-g3", it doesn't generate a ".debug_macro" section. This is because
the last option wins, thus downgrading the debug level again. Without
any macro debug information in the executable, info-macros.exp
obviously produces many failures.
Since the "-g" option is appended by DejaGnu's target_compile whenever
the "debug" option is set, the fix just removes that option.
gdb/testsuite/ChangeLog:
* gdb.base/info-macros.exp: Remove "debug" from the compile
options.
On 64-bit hosts unsigned long is 64 bit. Use uint32_t instead.
gdb/
2014-01-22 Baruch Siach <baruch@tkos.co.il>
* xtensa-tdep.h (xtensa_elf_greg_t): Change type to uint32_t.
Adds a section for --as-needed libraries to a linker map file, similar
to what we do for archive libraries.
bfd/
* elflink.c (elf_link_add_object_symbols): Call minfo for --as-needed.
ld/
* ldlang.c (asneeded_list_head, asneeded_list_tail): New vars.
(lang_init): Initialise them.
(lang_print_asneeded): New function.
(lang_process): Call lang_print_asneeded.
* ldlang.h (struct asneeded_minfo): New.
(asneeded_list_tail): Declare.
* ldmain.c (add_archive_element): Improve archive map heading.
* ldmisc.c (minfo): Stash --as-needed info.
The ARI script flagged the use of the __func__ variable, which
is normally not allowed (not defined in C90). However, this particular
use is OK, as the reference is only made when __STDC_VERSION__ >=
199901L. So, add an "ARI:" comment to explicitly OK this use.
gdb/ChangeLog:
* common/common-utils.h: Add "ARI:" comment beside __func__
reference.
While looking at this macro, I noticed that it wasn't always necessarily
defined. That prompted me to search the current sources to make sure
that all uses were adequately protected, which they were. But to help
prevent future uses to be made unprotected, this patch expands the
current macro documentation a bit.
gdb/ChangeLog:
* common/common-utils.h (FUNCTION_NAME): Expand the macro's
documentation a bit.
An assignment to dot in an output section that allocates space of
course keeps the output section. Here, I'm changing the behaviour for
assignments that don't allocate space. The idea is not so much to
allow people to force output of an empty section with ". = .", but
to fix cases where an otherwise empty section has padding added by an
alignment expression that changes with relaxation or .eh_frame
editing. Such a section might have zero size before relaxation and so
be stripped incorrectly.
ld/
* ld.texinfo (Output Section Discarding): Mention assigning to dot
as a way of keeping otherwise empty sections.
* ldexp.c (is_dot, is_value, is_sym_value, is_dot_ne_0,
is_dot_plus_0, is_align_conditional): New predicates.
(exp_fold_tree_1): Set SEC_KEEP when assigning to dot inside an
output section, except for some special cases.
* scripttempl/elfmicroblaze.sc: Use canonical form to align at
end of .heap and .stack.
ld/testsuite/
* ld-shared/elf-offset.ld: Align end of .bss with canonical form
of ALIGN that allows an empty .bss to be removed.
* ld-arm/arm-dyn.ld: Likewise.
* ld-arm/arm-lib.ld: Likewise.
* ld-elfvsb/elf-offset.ld: Likewise.
* ld-mips-elf/mips-dyn.ld: Likewise.
* ld-mips-elf/mips-lib.ld: Likewise.
* ld-arm/arm-no-rel-plt.ld: Remove duplicate ALIGN.
* ld-powerpc/vle-multiseg-1.ld: Remove ALIGN at start of section.
ALIGN address of section instead.
* ld-powerpc/vle-multiseg-2.ld: Likewise.
* ld-powerpc/vle-multiseg-3.ld: Likewise.
* ld-powerpc/vle-multiseg-4.ld: Likewise.
* ld-powerpc/vle-multiseg-6.ld: Likewise.
* ld-scripts/empty-aligned.d: Check section headers not program
headers. Remove xfail and notarget.
* ld-scripts/empty-aligned.t: Use canonical ALIGN for end of .text2.
After resolving a versioned reference, foo@VER1, to a default versioned
definition, foo@@VER1, from a shared object, we also merge it with
the existing regular default symbol definition, foo. When foo is IFUNC
and foo@@VER1 aren't, we will merge 2 incompatible definitions. This
patch avoids merging foo@@VER1 definition with foo definition if
one is IFUNC and the other isn't.
This is to avoid expressions like: %hi(foo) + 1, which will
not do what you expect. The complex relocations can handle it,
but the internal fixups can't.
gdb/
* configure.ac: Call AM_PROG_INSTALL_STRIP.
* configure: Regenerate.
* aclocal.m4: Regenerate.
* Makefile.in (install_sh, INSTALL_STRIP_PROGRAM, STRIP):
New substituted variables.
(install-strip): New target.
(INSTALL_SCRIPT): New substituted variable.
(FLAGS_TO_PASS): Add it.
(install-only): Use $(INSTALL_SCRIPT) rather than
$(INSTALL_PROGRAM) for gcore.
There is nothing linker can do when a type mismatched default definition
are made dynamic by info->shared, info->export_dynamic or h->ref_dynamic.
But we do want to avoid exporting it when building PIE. Let's remove
those checks.
bfd/
PR ld/2404
* elflink.c (_bfd_elf_merge_symbol): Don't check info->shared,
info->export_dynamic, nor !h->ref_dynamic for type mismatch when
adding the default version.
ld/testsuite/
PR ld/2404
* ld-elf/shared.exp: Add a PIE test for PR ld/2404.
