Make it clearer that the size field indicates the size of the section.
gdb/
* corelow.c (core_target::get_core_register_section): Rename
min_size to section_min_size.
This adds the target and native configure support, and the NEWS entries for
the new target and native configurations.
gdb/
* Makefile.in (ALL_TARGET_OBS): Add riscv-linux-tdep.c.
(ALLDEPFILES): Add riscv-linux-nat.c, and riscv-linux-tdep.c.
* NEWS: Mention new GNU/Linux RISC-V target.
* configure.host: Add riscv*-*-linux*.
* configure.nat: Add riscv*.
* configure.tgt: Add riscv*-*-linux*.
Make the infrun.c:resume function static, and update the header
comment on the infrun.c:proceed function. There should be no user
visible change after this commit.
gdb/ChangeLog:
* infrun.c (resume): Make static, add forward declaration.
(proceed): Update header comment.
* infrun.h (resume): Delete declaration.
This fixes some minor formatting issues in riscv-tdep.h, including one
pointed out by ARI.
ChangeLog
2018-08-09 Tom Tromey <tom@tromey.com>
* riscv-tdep.h: Minor formatting fixes.
ARI produces this warning for the lines touched in this patch:
warning: gettext: All messages should be marked up with _.
However, in these cases, the message is not translatable (they are
syscall names). Adding an extra set of parentheses silences the
warning.
gdb/ChangeLog:
* common/scoped_mmap.c (mmap_file): Silence ARI warning.
* dwarf-index-cache.c (create_dir_and_check): Likewise.
(test_mkdir_recursive): Likewise.
* dwarf-index-write.c (write_psymtabs_to_index): Likewise.
If a vla is not in memory, and the upper bound is not defined, then we
can't know that an array element exists or not, and we should not try
to access the array element. One case where this happens is for
arrays that have been optimised out, the array will then have
VALUE_LVAL of not_lval, and an undefined upper bound, if we then try
to access an element of this array we will index into random GDB
memory.
An argument could be made that even for arrays that are in inferior
memory, if the upper bound is not defined then we should not try to
access the array element, however, in some of the Fortran tests, it
seems as though we do rely indexing from a base address into an array
which has no bounds defined. In this case GDBs standard protection
for detecting unreadable target memory prevents bad thing happening.
gdb/ChangeLog:
* valarith.c (value_subscripted_rvalue): If an array is not in
memory, and we don't know the upper bound, then we can't know that
the requested element exists or not.
gdb/testsuite/ChangeLog:
* gdb.base/vla-optimized-out.exp: Add new test.
The three test scripts:
gdb/testsuite/gdb.base/vla-optimized-out.exp
gdb/testsuite/gdb.base/vla-optimized-out-o3.exp
gdb/testsuite/gdb.base/vla-optimized-out-o3-strict.exp
are all pretty similar, with differences in the compile flags used,
and some of the expected results.
Instead of maintaining 3 files, merge them into a single test script,
and use parameters to control the test behaviour.
gdb/testsuite/ChangeLog:
* gdb.base/vla-optimized-out-o3.exp: Delete.
* gdb.base/vla-optimized-out-o3-strict.exp: Delete.
* gdb.base/vla-optimized-out.exp: Extend to cover all of the
deleted tests.
Fix a typo and add a missing one.
gdb/ChangeLog:
* target.c (str_comma_list_concat_elem): Fix typo in comment.
(target_options_to_string): Add comment.
Some distros enable _FORTIFY_SOURCE by default, which caught a failure
to check the result of "write" in scoped_mmap-selftests.c. This patch
fixes the problem.
ChangeLog
2018-08-08 Tom Tromey <tom@tromey.com>
* unittests/scoped_mmap-selftests.c: Check result of "write".
This adds software single step support that is needed by the linux native port.
This is modeled after equivalent code in the MIPS port.
This also fixes a few bugs in the compressed instruction decode support. Some
instructions are RV32/RV64 specific, and this wasn't being checked. Also, a
few instructions were accidentally using the non-compressed is_* function.
This has been tested on a HiFive Unleashed running Fedora, by putting a
breakpoint on start, typing stepi, and then holding down the return key until
it finishes, and observing that I see everything I expect to see along the way.
There is a problem in _dl_addr where I get into an infinite loop, but it seems
to be some synchronization code that doesn't agree with single step, so I have
to find the end of the loop, put a breakpoint there, continue, and then single
step again until the end.
gdb/
* riscv-tdep.c (enum opcode): Add jump, branch, lr, and sc opcodes.
(decode_register_index_short): New.
(decode_j_type_insn, decode_cj_type_insn): New.
(decode_b_type_insn, decode_cb_type_insn): New.
(riscv_insn::decode): Add support for jumps, branches, lr, and sc. New
local xlen. Check xlen when decoding ambiguous compressed insns. In
compressed decode, use is_c_lui_insn instead of is_lui_insn, and
is_c_sw_insn instead of is_sw_insn.
(riscv_next_pc, riscv_next_pc_atomic_sequence): New.
(riscv_software_single_step): New.
* riscv-tdep.h (riscv_software_single_step): Declare.
This allows the function to be used from riscv OS files, which also need to
depend on XLEN size.
gdb/
* riscv-tdep.c (riscv_isa_xlen): Drop static.
* riscv-tdep.h (riscv_isa_xlen): Add extern declaration.
Consider the following GDB session:
(gdb) target extended-remote :2347
(gdb) file /path/to/exe
(gdb) set remote exec-file /path/to/exe
(gdb) set detach-on-fork off
(gdb) break breakpt
(gdb) run
# ... hits breakpoint
(gdb) info inferiors
Num Description Executable
* 1 process 17001 /path/to/exe
2 process 17002 /path/to/exe
(gdb) kill
(gdb) info inferiors
Num Description Executable
* 1 <null> /path/to/exe
2 process 17002 /path/to/exe
(gdb) target extended-remote :2348
../../src/gdb/thread.c:660: internal-error: thread_info* any_thread_of_process(int): Assertion `pid != 0' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Or, from bug PR gdb/18050:
(gdb) start
(gdb) add-inferior -exec /path/to/exe
(gdb) target extended-remote :2347
../../src/gdb/thread.c:660: internal-error: thread_info* any_thread_of_process(int): Assertion `pid != 0' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
The issue is calling target.c:dispose_inferior with a killed inferior in
the inferior list. This assertion is fixed in this commit.
The new test for this issue only runs on platforms that support
'detach-on-fork', and when using
'--target_board=native-extended-gdbserver'.
gdb/ChangeLog:
PR gdb/18050:
* target.c (dispose_inferior): Don't dispose of inferiors that are
already killed.
gdb/testsuite/ChangeLog:
PR gdb/18050:
* gdb.server/extended-remote-restart.c: New file.
* gdb.server/extended-remote-restart.exp: New file.
Commit 87d6a7aa93 ("Add DWARF index cache") broke
gdb.fortran/nested-funcs.exp. Because of the new "set index-cache"
command, the expression "set index = 42" now fails:
set index = 42
Undefined set index-cache command: "= 42". Try "help set index-cache".
(gdb) PASS: gdb.fortran/nested-funcs.exp: set index = 42
Fix it by changing it to "set variable index = 42". Also, use
gdb_test_no_output to confirm that it worked (since that particular test
wrongfully passed).
gdb/testsuite/ChangeLog:
* gdb.fortran/nested-funcs.exp: Replace "set index = 42" with
"set variable index = 42".
Return an std::string instead of a char *, saving some manual freeing.
I only manually tested with "set debug target 1" and "set debug lin-lwp
1", since this only deals with debug output.
gdb/ChangeLog:
* target.h (target_options_to_string): Return an std::string.
* target.c (str_comma_list_concat_elem): Return void, use
std::string.
(do_option): Likewise.
(target_options_to_string): Return an std::string.
* linux-nat.c (linux_nat_target::wait): Adjust.
* target-debug.h (target_debug_print_options): Adjust.
While looking into PR build/8751 (which seems to be fixed), I noticed
that it's not possible to change CPPFLAGS for gdb on the "make"
command line. It's reasonable to want to do this sometimes, and I
think this patch should suffice.
gdb/ChangeLog
2018-08-07 Tom Tromey <tom@tromey.com>
* Makefile.in (CPPFLAGS): New variable.
(INTERNAL_CPPFLAGS): Use it.
New in v3:
- Address Eli's comments.
This patch adds doc and news for the feature introduced by the previous
patch.
gdb/ChangeLog:
* NEWS: Mention the index cache.
gdb/doc/ChangeLog:
* gdb.texinfo (Index Files Speed Up GDB): Add section about
symbol index cache.
New in v3:
- Remove things related to the dwarf-5 format.
- Fix compilation on mingw (scoped_mmap.c).
GDB can generate indexes for DWARF debug information, which, when
integrated in the original binary, can speed up loading object files.
This can be done using the gdb-add-index script or directly by the
linker itself. However, not many people know about this. And even
among those who do, because it requires additional steps, I don't know a
lot of people who actually go through that trouble.
To help make using the DWARF index more transparent, this patch
introduces a DWARF index cache. When enabled, loading an index-less
binary in GDB will automatically save an index file in ~/.cache/gdb.
When loading that same object file again, the index file will be looked
up and used to load the DWARF index. You therefore get the benefit of
the DWARF index without having to do additional manual steps or
modifying your build system. When an index section is already present
in the file, GDB will prefer that one over looking up the cache.
When doing my edit-compile-debug cycle, I often debug multiple times the
same build, so the cache helps reducing the load time of the debug
sessions after the first one.
- The saved index file is exactly the same as the output of the "save
gdb-index" command. It is therefore the exact same content that would
be found in the .gdb_index or .debug_names section. We just leave it
as a standalone file instead of merging it in the binary.
- The cache is just a directory with files named after the object
file's build-id. It is not possible to save/load the index for an
object file without build-id in the cache.
- The cache uses the gdb index format. The problem with the dwarf-5
format is that we can generate an addendum to the .debug_str section
that you're supposed to integrate to the original binary. This
complicates a little bit loading the data from the cached index files,
so I would leave this for later.
- The size taken up by ~/.cache/gdb is not limited. I was thinking we
could add configurable limit (like ccache does), but that would come
after. Also, maybe a command to flush the cache.
- The cache is disabled by default. I think once it's been out there
and tested for a while, it could be turned on by default, so that
everybody can enjoy it.
- The code was made to follow the XDG specification: if the
XDG_CACHE_HOME environment variable, it is used, otherwise it falls
back to ~/.cache/gdb. It is possible to change it using "set
index-cache directory". On other OSes than GNU/Linux, ~/.cache may
not be the best place to put such data. On macOS it should probably
default to ~/Library/Caches/... On Windows, %LocalAppData%/... I
don't intend to do this part, but further patches are welcome.
- I think that we need to be careful that multiple instances of GDB
don't interfere with each other (not far fetched at all if you run GDB
in some automated script) and the cache is always coherent (either the
file is not found, or it is found and entirely valid). Writing the
file directly to its final location seems like a recipe for failure.
One GDB could read a file in the index while it is being written by
another GDB. To mitigate this, I made write_psymtabs_to_index write
to temporary files and rename them once it's done. Two GDB instances
writing the index for the same file should not step on each other's
toes (the last file to be renamed will stay). A GDB looking up a file
will only see a complete file or no file. Also, if GDB crashes while
generating the index file, it will leave a work-in-progress file, but
it won't be picked up by other instances looking up in the cache.
gdb/ChangeLog:
* common/pathstuff.h (get_standard_cache_dir): New.
* common/pathstuff.c (get_standard_cache_dir): New.
* build-id.h (build_id_to_string): New.
* dwarf-index-common.h (INDEX4_SUFFIX, INDEX5_SUFFIX,
DEBUG_STR_SUFFIX): Move to here.
* dwarf-index-write.c (INDEX4_SUFFIX, INDEX5_SUFFIX,
DEBUG_STR_SUFFIX): Move from there.
(write_psymtabs_to_index): Make non-static, add basename
parameter. Write to temporary files, rename when done.
(save_gdb_index_command): Adjust call to
write_psymtabs_to_index.
* dwarf2read.h (dwarf2_per_objfile) <index_cache_res>: New
field.
* dwarf2read.c (dwz_file) <index_cache_res>: New field.
(get_gdb_index_contents_from_cache): New.
(get_gdb_index_contents_from_cache_dwz): New.
(dwarf2_initialize_objfile): Read index from cache.
(dwarf2_build_psymtabs): Save to index.
* dwarf-index-cache.h: New file.
* dwarf-index-cache.c: New file.
* dwarf-index-write.h: New file.
gdb/testsuite/ChangeLog:
* boards/index-cache-gdb.exp: New file.
* gdb.dwarf2/index-cache.exp: New file.
* gdb.dwarf2/index-cache.c: New file.
* gdb.base/maint.exp: Check if we are using the index cache.
The following patch makes use of the mkdir function. Import the mkdir
gnulib module to ensure proper operation on all platforms.
gdb/ChangeLog:
* gnulib/aclocal.m4: Re-generate.
* gnulib/config.in: Re-generate.
* gnulib/configure: Re-generate.
* gnulib/import/Makefile.am: Re-generate.
* gnulib/import/Makefile.in: Re-generate.
* gnulib/import/m4/gnulib-cache.m4: Re-generate.
* gnulib/import/m4/gnulib-comp.m4: Re-generate.
* gnulib/import/m4/mkdir.m4: New file.
* gnulib/import/mkdir.c: New file.
* gnulib/update-gnulib.sh (IMPORTED_GNULIB_MODULES): Add mkdir
module.
New in v2:
- As Tom pointed out, we don't need to keep the fd around after
mmapping. This simplifies things quite a bit, since we don't need a
new class. It's now just a function that returns a scoped_mmap.
We already have scoped_mmap, which is a thin RAII layer over mmap. If
one simply wants to mmap an entire file for reading, it takes a bit of
boilerplate. This patch introduces the mmap_file function to make this
easier.
gdb/ChangeLog:
* Makefile.in (COMMON_SFILES): Add common/scoped_mmap.c.
* common/scoped_mmap.c: New file.
* common/scoped_mmap.h (destroy): New method.
(~scoped_mmap, reset): Use destroy.
(scoped_mmap): New move constructor.
(mmap_file): New declaration.
* unittests/scoped_mmap-selftests.c (test_normal,
test_invalid_filename, run_tests): New functions.
(_initialize_scoped_mmap_selftests): Register selftest.
New in v3:
- Remove changed to dwarf-5 functions.
The read_gdb_index_from_section and read_debug_names_from_section
functions read the index content, as their names state, from sections of
object files. A following patch will make it possible to read index
content from standalone files.
This patch therefore decouples the code that reads the index content
from the code that processes that content. Functions
dwarf2_read_gdb_index and dwarf2_read_debug_names receive callbacks that
are responsible for providing the index contents (for both the main file
and the potential dwz file).
gdb/ChangeLog:
* dwarf2read.c (read_gdb_index_from_section): Rename to...
(read_gdb_index_from_buffer): ... this. Remove section
parameter, add buffer parameter.
(get_gdb_index_contents_ftype,
get_gdb_index_contents_dwz_ftype): New typedefs.
(dwarf2_read_gdb_index): Add callback parameters to get the
index contents.
(get_gdb_index_contents_from_section): New.
(dwarf2_initialize_objfile): Update call to
dwarf2_read_gdb_index.
I noticed that make -jN check would run make check-single when using GNU
make 4.2.1.
In the end, it turned out that this is due to this change from the make
4.2 NEWS file:
* The amount of parallelism can be determined by querying MAKEFLAGS, even when
the job server is enabled (previously MAKEFLAGS would always contain only
"-j", with no number, when job server was enabled).
The fix is trivial: just accept an optional arg to -j in Makefile.in
(saw_dash_j). Tested on i386-pc-solaris2.11 with just make and make
-j/-jN with both make 3.82 and 4.2.1.
* Makefile.in (saw_dash_j): Allow for GNU make 4.2+ passing -jN in
MAKEFLAGS.
A recent patch introduced a few of these:
/home/emaisin/src/binutils-gdb/gdb/remote.c:12862:19: error: format not a string literal and no format arguments [-Werror=format-security]
error (err_msg);
^
Fix them by replacing the call to error with
error ("%s", err_msg);
gdb/ChangeLog:
* remote.c (remote_target::download_tracepoint): Fix format
string errors.
The tracefile.c:trace_save function assumes trace_regblock_size won't
be larger than the MAX_TRACE_UPLOAD constant, used to size the buffer
which holds trace data. This can cause buffer overruns when this is
not the case. This patch changes this function so that the larger
size is used to size the buffer.
gdb/ChangeLog:
2018-08-06 Pedro Franco de Carvalho <pedromfc@linux.ibm.com>
* tracefile.c: Include common/byte-vector.h.
(trace_save): Change type of buf to gdb::byte_vector. Initialize
with trace_regblock_size if needed. Update uses of buf.
This patch changes collection_list to allow larger register masks.
The mask is changed from an array to a vector and is initialized to
hold the maximum possible remote register number. The stringify
method is updated to resize temp_buf if needed.
gdb/ChangeLog:
2018-08-06 Pedro Franco de Carvalho <pedromfc@linux.ibm.com>
* tracepoint.h (collection_list) <m_regs_mask>: Change type to
std::vector<unsigned char>.
* tracepoint.c (collection_list::collection_list): Remove
m_regs_mask initializer from initializer list. Resize
m_regs_mask using the largest remote register number.
(collection_list::add_remote_register): Remove size check on
m_regs_mask. Use at to access element.
(collection_list::stringify): Change type of temp_buf to
gdb::char_vector. Update uses of temp_buf. Resize if needed to
stringify the register mask. Use pack_hex_byte for the register
mask.
Currently, tracepoint register masks in the QTDP packets include both
internal and remote register numbers, as well as pseudo-register
numbers.
This patch changes this so that the mask only includes remote register
numbers.
Register numbers from agent expressions are already set in the mask
using remote numbers. Other tracepoint actions used internal numbers,
e.g. "collect $regs" or "collect $<pseudoreg>". To handle pseudoreg
numbers, an empty agent expression is created and ax_reg_mask is
called for this expression and the pseudoreg. This will cause the ax
to set its mask with the corresponding remote raw register
numbers (using ax_regs_mask, which calls
gdbarch_ax_pseudo_register_collect).
If ax_regs_mask and gdbarch_ax_pseudo_register_collect also generate
more ax bytecode, the ax is also appended to the collection list. It
isn't clear that this was the original intent for
gdbarch_ax_pseudo_register_collect, and none of the arches seem to do
this, but if this changes in the future, it should work.
The patch also refactors the code used by validate_action line to
validate axs into a function that is now called from every place that
generates axs. Previously, some parts of tracepoint.c that generated
axs didn't check if the ax length was greater than MAX_AGENT_EXPR_LEN.
gdb/ChangeLog:
2018-08-06 Pedro Franco de Carvalho <pedromfc@linux.ibm.com>
* tracepoint.h (class collection_list) <add_register>: Remove.
<add_remote_register, add_ax_registers, add_local_register>:
Declare.
<add_memrange>: Add scope parameter.
* tracepoint.c (encode_actions_1): Likewise.
(collection_list::add_register): Rename to ...
(collection_list::add_remote_register): ... this. Update
comment.
(collection_list::add_ax_registers, add_local_register): New
methods.
(collection_list::add_memrange): Add scope parameter. Call
add_local_register instead of add_register.
(finalize_tracepoint_aexpr): New function.
(collection_list::collect_symbol): Update calls to add_memrange.
Call add_local_register instead of add_register. Call
add_ax_registers. Call finalize_tracepoint_aexpr.
(encode_actions_1): Get remote regnos for $reg action. Call
add_remote_register, add_ax_registers, and add_local_register.
Update call to add_memrange. Call finalize_tracepoint_aexpr.
(validate_actionline): Call finalize_tracepoint_aexpr.
This patch changes the remote target to use the remote packet size to
build QTDP packets, and to check if there is enough room for the
packet.
I changed the function to raise an error if the packet is too small,
instead of aborting gdb (through xsnprintf). It isn't clear if gdb
will be in a consistent state with respect to the stub after this,
since it's possible that some packets will be sent but not others, and
there could be an incomplete tracepoint on the stub.
The char array used to build the packets is changed to a
gdb::char_vector and sized with the result from
get_remote_packet_size.
When checking if the buffer is large enough to hold the tracepoint
condition agent expression, the length of the expression is multiplied
by two, since it is encoded with two hex digits per expression
byte. For simplicity, I assume that the result won't overflow, which
can happen for very long condition expressions.
gdb/ChangeLog:
2018-08-06 Pedro Franco de Carvalho <pedromfc@linux.ibm.com>
* remote.c (remote_target::download_tracepoint): Remove BUF_SIZE.
Replace array buf with gdb::char_vector buf, of size
get_remote_packet_size (). Replace references to buf and
BUF_SIZE to buf.data () and buf.size (). Replace strcpy, strcat
and xsnprintf with snprintf. Raise errors if the buffer is too
small.
The has_more predicate in remote_target::download_tracepoint always
evaluates to true, so the last action packet will be sent with a
trailing '-'. This patch changes the predicate to remove the last
trailing '-'.
gdb/ChangeLog:
2018-08-06 Pedro Franco de Carvalho <pedromfc@linux.ibm.com>
* remote.c (remote_target::download_tracepoint): Fix the has_more
predicate in the QTDP action list iteration.
gdb/ChangeLog:
2018-08-06 Pedro Franco de Carvalho <pedromfc@linux.ibm.com>
* remote.c (remote_target::download_tracepoint): Fix indentation
in for block.
When looking at the gdb.sum file produced by dg-extract-results.sh on
Solaris 11/x86, I noticed some wrong sorting, like this:
PASS: gdb.ada/addr_arith.exp: print something'address + 0
PASS: gdb.ada/addr_arith.exp: print 0 + something'address
PASS: gdb.ada/addr_arith.exp: print something'address - 0
PASS: gdb.ada/addr_arith.exp: print 0 - something'address
Looking closer, I noticed that while dg-extract-results.sh had been
copied over from contrib in the gcc repo, the corresponding
dg-extract-results.py file had not. The latter not only fixes the
sorting problem I'd observed, but is also way faster than the shell
version (like a factor of 50 faster).
Therefore I propose to update both files from the gcc repo. The changes
to the .sh version are trivial, just counting the number of DejaGnu
ERROR lines, too.
The files are moved to toplevel contrib:
* This way, they can easily be used should someone decide to parallelize
one or more of the binutils, gas, or ld testsuites.
* They are less easily overlooked for updates from the gcc repo when
they reside in the same place in both.
* The test_summary script needs to live in contrib since the toplevel
Makefile's mail-report.log target expects it there.
Tested on amd64-pc-solaris2.11 with
make -j16 check
and
make -j16 -k RACY_ITER=5 check
gdb/testsuite:
* dg-extract-results.sh: Move to toplevel contrib.
* Makefile.in (check-parallel): Reflect dg-extract-results.sh move.
* Makefile.in (check-parallel-racy): Likewise.
contrib:
* dg-extract-results.sh: Move from gdb/testsuite.
Update from gcc repo.
* dg-extract-results.py: New from gcc repo.
Not all targets support scheduler-locking. Add a check to see if the
taraget supports scheduler locking, and if it doesn't, don't run the
scheduler-locking tests that will otherwise fail.
There are actually a set of tests that try to use scheduler-locking
however, in most of these cases the test will not be run on smaller
targets (those that might not support threads and scheduler-locking)
due to the targets lack of support for threads, or some other larger
feature.
In the gdb.mi/mi-cmd-param-changed.exp test though, there's no
dependence on threads, or any other larger feature, and so, for the
small target I was using the test would otherwise try to run, only to
fail due to lack of support for scheduler-locking.
gdb/testsuite/ChangeLog:
* lib/gdb.exp (target_supports_scheduler_locking): New proc.
* gdb.mi/mi-cmd-param-changed.exp: Only run scheduler locking
tests if the target supports scheduler locking.
The MIPS target supports 127 signals, and this can create an ambiguity
in process wait statuses. A status value of 0x007f could potentially
indicate a process that has exited with signal 127, or a process that
has stopped with signal 0.
In uClibc-ng the interpretation of 0x007f is that the process has
exited with signal 127 rather than stopped with signal 0, and so,
WIFSTOPPED (W_STOPCODE (0)) will be false rather than true as it would
be on most other platforms.
Given that it's pretty easy to avoid using W_STOPCODE (0), lets do that.
gdb/ChangeLog:
* linux-nat.c (linux_nat_target::follow_fork): Avoid using
'W_STOPCODE (0)' as this could be ambiguous.
This fails for me on openSUSE leap 15.0:
...
FAIL: gdb.python/py-rbreak.exp: check number of returned breakpoints is 11
...
The rbreak "" command expects 11 breaks, but I see two extra for
__libc_csu_fini and __libc_csu_init:
...
Breakpoint 13 at 0x4005b0: file elf-init.c, line 106.^M
Breakpoint 14 at 0x400540: file elf-init.c, line 68.^M
...
This patch fixes the failing test by excluding functions starting with an
underscore.
Tested on x86_64-linux.
2018-08-04 Tom de Vries <tdevries@suse.de>
* gdb.python/py-rbreak.exp: Fix rbreak regexp.
This patch fixes a thinko that happened when I was implementing the
IPv6 support on GDB/gdbserver. On certain situations, it is necessary
to disable TCP's Nagle algorithm (NODELAY). For obvious reasons, this
only applies when we are dealing with a TCP connection.
While implementing the IPv6 patch, I noticed that the net_open
function (on gdb/ser-tcp.c) kept a flag indicating whether the
connection type was UDP or TCP. I eliminated that flag, and started
using the 'struct addrinfo *' related to the successful connection
directly. However, I made a mistake:
if (success_ainfo->ai_socktype == IPPROTO_TCP)
^^^^^^^^^^^
{
/* Disable Nagle algorithm. Needed in some cases. */
int tmp = 1;
setsockopt (scb->fd, IPPROTO_TCP, TCP_NODELAY,
(char *) &tmp, sizeof (tmp));
}
The 'ai_socktype' field specifies the socket type (SOCK_STREAM or
SOCK_DGRAM), and not the protocol. This test was always failing, and
the Nagle algorithm was never being disabled.
The obvious fix is to use the 'ai_protocol' field. This is what this
patch does.
Huge "thank you" to Joel Brobecker who reported the regression (he was
experiencing an unusual delay while debugging a bare-metal program
running under QEMU) and helped me set up a proper reproducer for the
bug.
gdb/ChangeLog:
2018-08-03 Sergio Durigan Junior <sergiodj@redhat.com>
* ser-tcp.c (net_open): Fix thinko when deciding whether to
disable TCP's Nagle algorithm (use "ai_protocol" instead of
"ai_socktype").
PR symtab/16842 shows that gdb will crash when the user tries to
invoke "info address" of a template parameter.
The bug here is that dwarf2read.c does not set the symtab on the
template parameter symbols. This is pedantically correct, given that
the template symbols do not appear in a symtab. However, gdb
primarily uses the symtab backlink to find the symbol's objfile. So,
this patch simply sets the symtab on these symbols.
Tested by the buildbot.
gdb/ChangeLog
2018-08-02 Tom Tromey <tom@tromey.com>
PR symtab/16842.
* dwarf2read.c (read_func_scope): Set symtab on template parameter
symbols.
(process_structure_scope): Likewise.
gdb/testsuite/ChangeLog
2018-08-02 Tom Tromey <tom@tromey.com>
PR symtab/16842.
* gdb.cp/temargs.exp: Test "info address" of a template
parameter.
Starting with MacOS version Sierra, the gdb kill command
seems to work but inferior remains as zombie on the host.
Notice that, as zombie process, the inferior is not killable
by the user, nor by root.
The kill signal gdb sent to the inferior is not handled
in gdb as a signal sent by gdb thus no reply is made and
the process remains (since MacOS does not "release" the
inferior because no reply have been made to the signal
message).
This patch fixes this problem.
gdb/ChangeLog
2018-08-02 Xavier Roirand <roirand@adacore.com>
PR gdb/22629:
* darwin-nat.c (darwin_kill_inferior): Fix handling of
kill inferior.
I noticed that the existing kill-detach-inferiors-cmd.exp test was
causing gdb to crash on macOS 10.13. The bug was that an inferior
that hadn't yet been started would cause get_darwin_inferior to return
NULL, and this was not checked.
I went through the places calling get_darwin_inferior and added checks
where appropriate. This makes the test get a bit further. Not all of
these spots are exercised by the test, but they seem safe enough in
any case.
gdb/ChangeLog
2018-08-02 Tom Tromey <tom@tromey.com>
* darwin-nat.c (find_inferior_task_it, darwin_find_thread)
(darwin_suspend_inferior, darwin_resume_inferior)
(darwin_decode_notify_message, darwin_resume_inferior_threads)
(darwin_check_new_threads): Check result of get_darwin_inferior.
Invoking -var-info-path-expression on a dynamic varobj lead either in wrong
(nonsense) result or to a segmentation fault in cplus_describe_child().
This was caused by the fact that varobj_get_path_expr() called
cplus_path_expr_of_child() ignoring the fact the parent of the variable
is dynamic. Then, cplus_describe_child() accessed the underlaying C type
members by index, causing (i) either wrong (nonsense) expression being
returned (since dynamic child may be completely arbibtrary value)
or (ii) segmentation fault (in case the index higher than number of
underlaying C type members.
This fixes the problem by checking whether a varobj is a child of a dynamic
varobj and, if so, reporting an error as described in documentation.
gdb/ChangeLog:
* varobj.c (varobj_get_path_expr_parent): Report an error if
parent is a dynamic varobj.
gdb/testsuite/Changelog:
* gdb.python/py-mi-var-info-path-expression.c: New file.
* gdb.python/py-mi-var-info-path-expression.py: New file.
* gdb.python/py-mi-var-info-path-expression.exp: New file.
I noticed that re-generating our gnulib import introduced some changes.
I supposed that this comes from the recent upgrade to autoconf 2.69
(though I haven't checked).
Tested by rebuilding on GNU/Linux x86-64 and mingw (cross-compiled from
GNU/Linux).
gdb/ChangeLog:
* gnulib/aclocal.m4: Re-generate.
* gnulib/config.in: Re-generate.
* gnulib/configure: Re-generate.
* gnulib/import/Makefile.in: Re-generate.
* gnulib/import/m4/gnulib-comp.m4: Re-generate.
* gnulib/import/m4/onceonly.m4: Re-generate.
Looking at the address sanitizer output, this was a quite low hanging
fruit. We create target_desc objects for testing that we never free.
Saving them in unique_ptrs takes care of it.
I created a small struct to hold these because I thought it would help
readability.
gdb/ChangeLog:
* target-descriptions.c (struct xml_test_tdesc): New.
(xml_tdesc): Change type to std::vector<xml_test_tdesc>.
(record_xml_tdesc): Update.
(maintenance_check_xml_descriptions): Update.
* target-descriptions.h (record_xml_tdesc): Update comment.
On commit:
commit 7f1f7e2393
Author: Sergio Durigan Junior <sergiodj@redhat.com>
Date: Fri Jul 13 16:20:34 2018 -0400
Expect for another variant of error message when gdbserver cannot resolve hostname
I extended the regular expression being used to identify whether
gdbserver could not resolve a (host)name. This was needed because the
error message being printed had a different variation across some
systems. However, as it turns out, I've just noticed that the message
has yet another variation:
target remote tcp8:123:2353
tcp8:123:2353: cannot resolve name: System error
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
tcp8:123:2353: No such file or directory.
(gdb) FAIL: gdb.server/server-connect.exp: tcp8: connect to gdbserver using tcp8:123
which is causing FAILs on some systems (namely, Fedora-i686 on
BuildBot).
So instead of trying to predict everything that can be printed, I
decided to just match anything after the "cannot resolve name: " part.
This patch implements that.
Regression tested on the BuildBot.
gdb/testsuite/ChangeLog:
2018-07-30 Sergio Durigan Junior <sergiodj@redhat.com>
* lib/gdbserver-support.exp (gdbserver_start): Match any kind of
error after "cannot resolve name" string.
In commit:
commit 37cc0caeca
Date: Wed Jul 18 13:38:35 2018 +0200
[gdb/exp] Interpret size of vla with unknown size as <optimized out>
All dynamic types are treated as arrays in the 'sizeof' code path,
which means that structures can incorrectly be treated as arrays.
This can cause a failure in the gdb.base/vla-datatypes.exp test
script.
This commit adds a check that we do have an array before checking the
array bounds, and I also check that the array index type is dynamic
too. This second check probably isn't strictly necessary, but
shouldn't hurt, a non-dynamic index type shouldn't have undefined high
bound.
gdb/ChangeLog:
* eval.c (evaluate_subexp_for_sizeof): Check for array type before
checking array bounds are defined.
I noticed a buildbot failure where gdb crashed in info-os.exp, when
compiled with -D_GLIBCXX_DEBUG:
(gdb) info os procgroups
/usr/include/c++/7/bits/stl_algo.h:4834:
Error: comparison doesn't meet irreflexive requirements, assert(!(a < a)).
Objects involved in the operation:
iterator::value_type "< operator type" {
type = pid_pgid_entry;
}
The bug here is that pid_pgid_entry::operator< violates the C++
irreflexivity rule; that is, that an object cannot be less than
itself.
Tested locally by re-running info-os.exp.
gdb/ChangeLog
2018-07-30 Tom Tromey <tom@tromey.com>
* nat/linux-osdata.c (pid_pgid_entry::operator<): Fix
irreflexivity violation.
This removes dead code that, according to the comments, existed to
placate lint. I don't think this has been relevant in a long time,
and certainly not since gdb switched to C++.
Tested by rebuilding.
ChangeLog
2018-07-30 Tom Tromey <tom@tromey.com>
* cli/cli-decode.c (lookup_cmd): Remove lint code.
* value.c (unpack_long): Remove lint code.
* valops.c (value_ind): Remove lint code.
* valarith.c (value_x_binop, value_x_unop, value_equal)
(value_pos): Remove lint code.
The gdb_load_shlib function will, on remote targets, try to run some
GDB commands. This obviously isn't going to work unless GDB is
running.
The gdb.trace/tspeed.exp test calls gdb_load_shlib before starting
GDB. Don't do that.
The failure that's triggered is actually DeJaGNU complaining that the
variable $use_gdb_stub doesn't exist, this is only created when GDB is
started. Something like this should trigger a failure:
make check-gdb \
RUNTESTFLAGS="--target_board=remote-gdbserver-on-localhost \
gdb.trace/tspeed.exp"
This commit also adds a check to gdb_load_shlib that GDB is running.
The check is always performed, so this should catch cases where a GDB
developer adds a use of gdb_load_shlib but doesn't test their code
with a remote target.
gdb/testsuite/ChangeLog:
* gdb.trace/tspeed.exp: Only call gdb_load_shlib after gdb has
started.
* lib/gdb.exp (gdb_load_shlib): Call perror if GDB is not running.
Alan Hayward
Jim Wilson
Andrew Burgess
Tom Tromey
Simon Marchi
Szabolcs Nagy
Simon Marchi
Rainer Orth
Pedro Franco de Carvalho
Tom de Vries
Sergio Durigan Junior
Xavier Roirand
Joel Brobecker
Jan Vrany