I noticed in gdb.threads/threadcrash.exp that the usage of test_list is
somewhat convoluted.
Simplify the test-case by storing a classification instead of a pattern in
test_list.
Tested on arm-linux and x86_64-linux.
A linaro PR [1] reports that the gdb.threads/threadcrash.exp test-case fails
to cout the number of threads in the inferior:
...
FAIL: gdb.threads/threadcrash.exp: test_gcore: $thread_count == 7
FAIL: gdb.threads/threadcrash.exp: test_gcore: $thread_count == [llength $test_list]
...
Fix this by getting the convenience variable _inferior_thread_count as opposed
to calculating it based on the output of "info threads".
Tested on arm-linux and x86_64-linux.
Reviewed-By: Lancelot Six <lancelot.six@amd.com>
Approved-By: Tom de Vries <tdevries@suse.de>
[1] https://linaro.atlassian.net/browse/GNU-1120
With check-readmore, I run into:
...
FAIL: gdb.threads/threadcrash.exp: test_corefile: \
$thread_count == [llength $test_list]
...
The problem is that the clauses in the gdb_test_multiple for
"thread apply all backtrace" intent to match one line, but actually can
match more than one line, and consequently a match for one type of thread can
consume a line that was supposed to match another thread.
For instance, there's this regexp:
...
-re "\[^\n\]*syscall_task .location=SIGNAL_ALT_STACK\[^\n\]*" {
...
It's limited at the end by \[^\n\]*, meaning the match stops at the end of the
line.
But it doesn't start with a ^, and consequently can match more than one line.
The "\[^\n\]*" at the start doesn't prevent this, there's an implicit .* at
the start of each pattern, unless it's anchored using a ^.
Fix this by rewriting the regexps in a "^\r\n$hs$regexp$hs$eol" style, where:
- hs is: \[^\n\]* (horizontal space), and
- eol is (?=\r\n) (look-ahead end-of-line).
It also turned out to be necessary to drop the -lbl switch, and introduce a
corresponding explicit clause. The -lbl clause is placed ALAP, and
consequently allowed the default fail clause to trigger.
Tested on arm-linux and x86_64-linux.
In test-case gdb.threads/threadcrash.exp we have an unnecessarily indented
gdb_test_multiple:
...
gdb_test_multiple "thread apply all backtrace" \
"Get thread information" -lbl {
-re "#\[0-9\]+\\\?\\\?\[^\n\]*" {
...
Fix this by moving the command into a variable, allowing the
"gdb_test_multiple ... {" to fit on a single 80 chars line.
Tested on arm-linux and x86_64-linux.
Some of these have no explicit %xmm operand(s), yet they still act SSE-
like (in leaveing bits 128 and up untouched). Hence they want similarly
diagnosing, if that was asked for.
These aren't useful, but can be encoded for their AVX forms and hence
should also be permitted for the APX surrogates. Extend the respective
conditional by a base opcode check, to restrict it to VROUND{P,S}{S,D}.
Since ar can be built defaulting to deterministic mode, tests which
expect non-deterministic behaviour need to explicitly set the U flag.
The non-deterministic member test expects SOURCE_DATE_EPOCH to not be
set; this documents that. Unconditionally unsetting the variable
causes issues in test infrastructure (which expects unsetenv to only
be called on variables which are already set).
Signed-off-by: Stephen Kitt <steve@sk2.org>
Starting python version 3.12, PyErr_Fetch and PyErr_Restore are deprecated.
Use PyErr_GetRaisedException and PyErr_SetRaisedException instead, for
python >= 3.12.
Tested on aarch64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
With python 3.12, I run into:
...
(gdb) PASS: gdb.python/py-block.exp: check variable access
python print (block['nonexistent'])^M
Python Exception <class 'KeyError'>: 'nonexistent'^M
Error occurred in Python: 'nonexistent'^M
(gdb) FAIL: gdb.python/py-block.exp: check nonexistent variable
...
The problem is that that PyErr_Fetch returns a normalized exception, while the
test-case matches the output for an unnormalized exception.
With python 3.6, PyErr_Fetch returns an unnormalized exception, and the
test passes.
Fix this by:
- updating the test-case to match the output for a normalized exception, and
- lazily forcing normalized exceptions using PyErr_NormalizeException.
Tested on aarch64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
Similar to gdbpy_err_fetch::value, add a getter gdbpy_err_fetch::type, and use
both consistently to get gdbpy_err_fetch members m_error_value and
m_error_type.
Tested on aarch64-linux.
* bfd.c (_bfd_print): Renamed from bfd_print_error.
(bfd_print_error): Reinstate previous code but using the above.
(error_handler_fprintf, error_handler_sprintf): Adjust.
* bfd-in2.h: Regenerate.
Commit b1c95bc4dd cleared some bfd static variables, with bad
results since bfd_set_error_program_name is often called before
bfd_init.
* bfd.c (bfd_init): Don't clear _bfd_error_program_name.
* bfd.c (bfd_print_error): Make static. Don't print program name.
(error_handler_fprintf): Print program name here.
* format.c (print_warnmsg): Use _bfd_error_handler to print
cached messages.
* bfd-in2.h: Regenerate.
The background DWARF reader changes introduced a race when writing to
the index cache. The problem here is that constructing the
index_cache_store_context object should only happen on the main
thread, to ensure that the various value captures do not race.
This patch adds an assert to the construct to that effect, and then
arranges for this object to be constructed by the cooked_index_worker
constructor -- which is only invoked on the main thread.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31262
I think it is cleaner for 'store' to be a method on
index_cache_store_context rather than on the global index cache
itself. This patch makes this change.
This changes index_cache_store_context to also capture the per-BFD
object when it is constructed. This is used when storing to the
cache, and this approach makes the code a little simpler.
I noticed that index_cache_store_context captures the 'enabled'
setting, but not the index cache directory. This patch makes this
change, which avoids a possible race -- with background reading, the
user could possibly change this directory at the exact moment the
writer examines the variable.
A bug report in the DAP specification repository pointed out that it
is typical for DAP implementations to put a function's return value
into the outermost scope.
This patch changes gdb to follow this convention.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31341
Reviewed-By: Kévin Le Gouguec <legouguec@adacore.com>
This patch changes the Python "stop" event emission code to also add
the function return value, if it is known. This happens when the stop
comes from a "finish" command and when the value can be fetched.
The test is in the next patch.
Reviewed-By: Eli Zaretskii <eliz@gnu.org>
Add "()" to silence GCC 6.4:
.../gas/config/tc-i386.c: In function ‘x86_ginsn_lea’:
.../gas/config/tc-i386.c:5738:19: error: logical not is only applied to the left hand side of comparison [-Werror=logical-not-parentheses]
if (!i.base_reg != (!i.index_reg || i.index_reg->reg_num == RegIZ))
^~
cc1: all warnings being treated as errors
PR gas/31464
* config/tc-i386.c (x86_ginsn_lea): Add "()" to silence GCC 6.4.
PR gdb/31260 points out a race introduced by the background reading
changes. If a given objfile is re-opened when it is already being
read, dwarf2_initialize_objfile will call dwarf2_read_dwz_file again,
causing the 'dwz_file' to be reset.
This patch fixes the problem by arranging to open the dwz just once:
when the dwarf2_per_bfd object is created.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31260
This modification mainly changes the timing of type transition,
adds relaxation to the old LE instruction sequence, and fixes
bugs in extreme code models.
We strictly distinguish between type transition and relaxation.
Type transition is from one type to another, while relaxation
is the removal of instructions under the same TLS type. Detailed
instructions are as follows:
1. For type transition, only the normal code model of DESC/IE
does type transition, and each relocation is accompanied by a
RELAX relocation. Neither abs nor extreme will do type transition,
and no RELAX relocation will be generated.
The extra instructions when DESC transitions to other TLS types
will be deleted during the type transition.
2. Implemented relaxation for the old LE instruction sequence.
The first two instructions of LE's 32-bit and 64-bit models
use the same relocations and cannot be distinguished based on
relocations. Therefore, for LE's instruction sequence, any code
model will try to relax.
3. Some function names have been adjusted to facilitate understanding,
parameters have been adjusted, and unused macros have been deleted.
We don't want to lose an abort message when bfd_set_error_handler has
been called to ignore or cache errors.
PR ld/31444
* bfd.c (_bfd_abort): Don't use _bfd_error_handler.
This fixes some duplicate test names in gdb.trace/circ.exp when using
native-gdbserver and native-extended-gdbserver boards.
In this test we set the trace buffer size twice. The same test name
was used each time the size was adjusted.
I've fixed this issue by:
1. Creating a new proc, set_trace_buffer_size, which factors out the
code to change the buffer size, and uses test names based on the
size we're setting the buffer too,
2. Calling the new proc each time we want to adjust the buffer size.
After this the duplicate test names are resolved. There should be no
change in what is tested after this commit.
This commit fixes some duplicate test names in the gdb.trace/
directory when run with the native-gdbserver and
native-extended-gdbserver boards. In this case the duplicates relate
to the calls to gdb_compile_pthreads which emits a fixed PASS message,
as there are two calls to gdb_compile_pthreads we get a duplicate PASS
message.
In both cases the problem is fixed by adding a with_test_prefix around
one of the compilations, however, I've made additional changes to
clean up the tests a little while I was working on them:
1. Switch to use prepare_for_testing instead of
gdb_compile_pthreads. By passing the 'pthreads' option this does
call gdb_compile_pthreads under the hood, but using the standard
compile function is cleaner,
2. Using prepare_for_testing removes the need to call clean_restart
immediately afterwards, so those calls are removed,
3. I removed the unneeded $executable and $expfile globals, where
the $executable global was used I've replaced this with $binfile,
4. When we compile two executables I've now given these different
names so that both exist at the end of the test run,
5. Removed a gdb_reinitialize_dir call, this is covered by
clean_restart,
6. Use gdb_test_no_output where it makes sense.
I now see no duplicate test names when running these test scripts.
There should be no change in what is being tested after this commit.
There is an assertion error "gdb_assert (n < tdesc->reg_defs.size ())"
in find_register_by_number() when gdb connects to gdbserver, this
is because the value of LOONGARCH_LINUX_NUM_GREGSET (45, which contains
10 reserved regs) is different with the number of regs (35, which not
contains 10 reserved regs) in file gdb/features/loongarch/base64.xml.
Add a new macro LOONGARCH_USED_NUM_GREGSET which is defined as 35 to
keep consistent with the gdb/features/loongarch/base64.xml, and then
define LOONGARCH_FIRST_FP_REGNUM as LOONGARCH_USED_NUM_GREGSET so that
all the reg numbers in regcache are consistent with tdesc reg numbers.
without this patch:
Execute on the target machine:
$ gdbserver 192.168.1.123:5678 ./test
Execute on the host machine:
$ gdb ./test
(gdb) target remote 192.168.1.123:5678
Output on the target machine:
Process ./test created; pid = 67683
Listening on port 5678
Remote debugging from host 192.168.1.136, port 6789
gdbserver/regcache.cc:205: A problem internal to GDBserver has been detected.
find_register_by_number: Assertion 'n < tdesc->reg_defs.size ()' failed.
Output on the host machine:
Remote debugging using 192.168.1.123:5678
Remote connection closed
Signed-off-by: Hui Li <lihui@loongson.cn>
Approved-By: John Baldwin <jhb@FreeBSD.org>
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
