There are some subtle differences between architectures, like the size
of a "long" type, and this isn't currently accounted for in
gdb.base/parse_number.exp.
For example, on aarch64 a long type is 8 bytes, whereas a long type is
4 bytes for x86_64. This causes the following FAIL's:
FAIL: gdb.base/parse_number.exp: lang=asm: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=auto: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=c: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=c++: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=fortran: p/x 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=fortran: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=go: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=local: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=minimal: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=objective-c: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=opencl: ptype 0xffffffffffffffff
FAIL: gdb.base/parse_number.exp: lang=pascal: ptype 0xffffffffffffffff
There are some fortran-specific divergences as well, where 32-bit
architectures show "unsigned int" for both 32-bit and 64-bit integers
and 64-bit architectures show "unsigned int" and "unsigned long" for
32-bit and 64-bit integers.
There might be a bug that 32-bit fortran truncates 64-bit values to
32-bit, given "p/x 0xffffffffffffffff" returns "0xffffffff".
Here's what we get for aarch64:
(gdb) ptype 0xffffffff
type = unsigned int
(gdb) ptype 0xffffffffffffffff
type = unsigned long
(gdb) p sizeof (0xffffffff)
$1 = 4
(gdb) p sizeof (0xffffffffffffffff)
quit
$2 = 8
(gdb) ptype 0xffffffff
type = unsigned int
(gdb) ptype 0xffffffffffffffff
type = unsigned long
And for arm:
(gdb) ptype 0xffffffff
type = unsigned int
(gdb) ptype 0xffffffffffffffff
quit
type = unsigned long long
(gdb) p sizeof (0xffffffff)
quit
$1 = 4
(gdb) p sizeof (0xffffffffffffffff)
quit
$2 = 8
(gdb) ptype 0xffffffff
type = unsigned int
(gdb) ptype 0xffffffffffffffff
type = unsigned long
This patch...
* Makes the testcase iterate over all architectures, thus covering all
the different combinations of types/sizes every time.
* Adjusts the expected values and types based on the sizes of long
long, long and int.
A particularly curious architecture is s12z, which has 32-bit long
long, and thus no way to represent 64-bit integers in C-like
languages.
Co-Authored-By: Luis Machado <luis.machado@arm.com>
Change-Id: Ifc0ccd33e7fd3c7585112ff6bebe7d266136768b
I broke the gdbserver build on x86-64 Windows a little while back.
Previously, I could not build this configuration, but today I found
out that if I configure with:
--host=x86_64-w64-mingw32 --target=x86_64-w64-mingw32
using the Fedora 34 tools, it will in fact build. I'm not certain,
but maybe the gnulib update helped with this.
This patch fixes the build. I'm checking it in.
This patch makes use of the support for several stack pointers
introduced by the previous patch to switch between them as needed
during unwinding.
It introduces a new 'unwind-secure-frames' arm command to enable/disable
mode switching during unwinding. It is enabled by default.
It has been tested using an STM32L5 board (with cortex-m33) and the
sample applications shipped with the STM32Cube development
environment: GTZC_TZSC_MPCBB_TrustZone in
STM32CubeL5/Projects/NUCLEO-L552ZE-Q/Examples/GTZC.
The test consisted in setting breakpoints in various places and check
that the backtrace is correct: SecureFault_Callback (Non-secure mode),
__gnu_cmse_nonsecure_call (before and after the vpush instruction),
SecureFault_Handler (Secure mode).
This implies that we tested only some parts of this patch (only MSP*
were used), but remaining parts seem reasonable.
Signed-off-by: Torbjörn Svensson <torbjorn.svensson@st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@foss.st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@arm.com>
Armv8-M architecture with Security extension features four stack pointers
to handle Secure and Non-secure modes.
This patch adds support to switch between them as needed during
unwinding, and replaces all updates of cache->prev_sp with calls to
arm_cache_set_prev_sp.
Signed-off-by: Torbjörn Svensson <torbjorn.svensson@st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@foss.st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@arm.com>
This patch is a preparation for the rest of the series and adds two
arm_cache_init helper functions. It updates every place that updates
cache->saved_regs to call the helper instead.
Signed-off-by: Torbjörn Svensson <torbjorn.svensson@st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@foss.st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@arm.com>
This patch removes the hardcoded access to PSP in
arm_m_exception_cache() and relies on the definition with the XML
descriptions.
Signed-off-by: Christophe Lyon <christophe.lyon@foss.st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@arm.com>
While working on adding support for Non-secure/Secure modes unwinding,
I noticed that the prologue analysis lacked support for vpush, which
is used for instance in the CMSE stub routine.
This patch updates thumb_analyze_prologue accordingly, adding support
for vpush of D-registers.
Signed-off-by: Christophe Lyon <christophe.lyon@foss.st.com>
Signed-off-by: Christophe Lyon <christophe.lyon@arm.com>
Tom and Simon feedback that there is a test failing in this commit:
commit a5c69b1e49
Date: Sun Apr 17 15:09:46 2022 +0800
gdb: fix using clear command to delete non-user breakpoints(PR cli/7161)
Then, I reproduced the same fail with Ubuntu 20.04 as Simon said, and I
fixed the nit in this patch. The root of the problem is not correctly
matching the presentation of internal breakpoints.
In addition, as Pedro pointed out, the original testcase is not portable
in some methods, so this patch fixes this issue and some other
improvements.
Tested on x86_64 ubuntu 20.04.4 and openSUSE Tumbleweed(VERSION_ID="20220425").
Revert commit 65daf5bed6 testsuite changes in ld-plugin/. -z isn't
supported for non-ELF targets, and isn't needed since we now prune the
exec stack warning (commit 333cd559ba).
PR 29072
I learned about with_cwd today. I spotted a few spots that could use
it, to make the code more robust.
Change-Id: Ia23664cb827f25e79d31948e0c006a8dc61c33e1
This patch adds PowerPC specific tests to verify recording of various
instructions. The first test case checks the ISA 2.06 lxvd2x instruction.
The second test case tests several of the ISA 3.01 instructions. Specifically,
it checks the word and prefixed instructions and some of the Matrix
Multiply Assist (MMA) instructions.
The patch has been run on both Power 10 and Power 9 to verify the ISA
2.06 test case runs on both platforms without errors. The ISA 3.1 test
runs without errors on Power 10 and is skipped as expected on Power 9.
This patch adds support for the PowerPC ISA 3.1 instructions to the PowerPC
gdb instruction recording routines. Case statement entries are added to a
number of the existing routines for recording the 32-bit word instructions.
A few new functions were added to handle the new word instructions. The 64-bit
prefix instructions are all handled by a set of new routines. The function
ppc_process_prefix_instruction() is the primary function to handle the
prefixed instructions. It calls additional functions to handle specific
sets of prefixed instructions. These new functions are:
ppc_process_record_prefix_vsx_d_form(),
ppc_process_record_prefix_store_vsx_ds_form(),
ppc_process_record_prefix_op34(),
ppc_process_record_prefix_op33(),
ppc_process_record_prefix_op32(),
ppc_process_record_prefix_store(),
ppc_process_record_prefix_op59_XX3(),
ppc_process_record_prefix_op42().
Internally at AdaCore, we noticed that the new Windows thread name
code could fail. First, it might return a zero-length string, but in
gdb conventions it should return nullptr instead. Second, an encoding
failure could wind up showing replacement characters to the user; this
is confusing and not useful; it's better to recognize such errors and
simply discard the name. This patch makes both of these changes.
I noticed that these files failed to format with Black, because they use
print without parenthesis (which isn't Python 3 compatible).
I don't know if these files are still relevant, but the change is
trivial, so here it is.
Change-Id: I116445c2b463486016f824d32effffc915b60766
The format for printing the floating point values was changed by commit:
commit 56262a931b
Author: Tom Tromey <tromey@adacore.com>
Date: Thu Feb 17 13:43:59 2022 -0700
Change how "print/x" displays floating-point value
Currently, "print/x" will display a floating-point value by first
casting it to an integer type. This yields weird results like:
(gdb) print/x 1.5
$1 = 0x1
...
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=16242
The above change results in 417 regression test failures since the expected
Power vector register output no longer match.
This patch updates the expected Altivec floating point register prints to
the hexadecimal format for both big endian and little endian systems. The
patch also fixes a formatting isue with the decimal_vector expected value
assign statements.
The expected VSX vector_register1, vector_register1_vr, vector_register2,
vector_register2_vr variables are updated to include the new float128 entry.
Additionally, the comment in the vsx expect file about the initialization
of the vs registers is updated.
The patch has been tested on Power 10, Power 8 LE and Power 8 BE.
This fixes build breakage using clang with libc++ on FreeBSD where
std::array<> is not yet declared when used by the path_join variadic
function template.
PR 29072
bfd * elflink.c (bfd_elf_size_dynamic_sections): Display a note to the
user that the current ehaviour of creating an executable stack
because of a missing .note.GNU-stack section is deprecated and
will be changed in a future release.
binutils* testsuite/lib/binutils-common.exp (prune_warnings_extra): Filter
out notes about the executable stacjk behaviour beign deprecated.
ld * testsuite/ld-elf/pr29072.b.warn: Update to include the note
about the linker's behaviour being depreccated.
If a variable is passed to function in FORTRAN as an argument the
variable is treated as an array with rank zero. GDB currently does
not support the case for assumed rank 0. This patch provides support
for assumed rank 0 and updates the testcase as well.
Without patch:
Breakpoint 1, arank::sub1 (a=<error reading variable:
failed to resolve dynamic array rank>) at assumedrank.f90:11
11 PRINT *, RANK(a)
(gdb) p a
failed to resolve dynamic array rank
(gdb) p rank(a)
failed to resolve dynamic array rank
With patch:
Breakpoint 1, arank::sub1 (a=0) at assumedrank.f90:11
11 PRINT *, RANK(a)
(gdb) p a
$1 = 0
(gdb) p rank(a)
$2 = 0
While working in gdb/infrun.c:restart_threads, I was wondering what are
the preconditions to call the function. It seems to me that
!step_over_info_valid_p should be a precondition (i.e. if we are doing
an inline step over breakpoint, we do not want to resume non stepping
threads as one of them might miss the breakpoint which is temporally
disabled).
To convince myself that this is true, I have added an assertion to
enforce this, and got one regression in the testsuite:
FAIL: gdb.base/step-over-syscall.exp: vfork: displaced=off: single step over vfork (GDB internal error)
This call to restart_threads originates from handle_vfork_done which
does not check if a step over is active when restarting other threads:
if (target_is_non_stop_p ())
{
scoped_restore_current_thread restore_thread;
insert_breakpoints ();
restart_threads (event_thread, event_thread->inf);
start_step_over ();
}
In this patch, I propose to:
- Call start_step_over before restart_threads. If a step over is already
in progress (as it is the case in the failing testcase),
start_step_over return immediately, and there is no point in restarting
all threads just to stop them right away for a step over breakpoint.
- Only call restart_threads if no step over is in progress at this
point.
In this patch, I also propose to keep the assertion in restart_threads
to help enforce this precondition, and state it explicitly.
I have also checked all other places which call restart_threads, and
they all seem to check that there is no step over currently active
before doing the call.
As for infrun-related things, I am never completely sure I did not miss
something. So as usual, all feedback and thoughts are very welcome.
Tested on x86_64-linux-gnu.
Change-Id: If5f5f98ec4cf9aaeaabb5e3aa88ae6ffd70d4f37
After this commit:
commit d08cbc5d32
Date: Wed Dec 22 12:57:44 2021 +0000
gdb: unbuffer all input streams when not using readline
Issues were reported with some MS-Windows hosts, see the thread
starting here:
https://sourceware.org/pipermail/gdb-patches/2022-March/187004.html
Filed in bugzilla as: PR mi/29002
The problem seems to be that calling setbuf on terminal file handles
is not always acceptable, see this mail for more details:
https://sourceware.org/pipermail/gdb-patches/2022-April/187310.html
This commit does two things, first moving the setbuf calls out of
gdb_readline_no_editing_callback so that we don't end up calling
setbuf so often.
Then, for MS-Windows hosts, we don't call setbuf for terminals, this
appears to resolve the issues that have been reported.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29002
When running:
$ make check TESTS="gdb.multi/multi-re-run.exp" RUNTESTFLAGS="--target_board=native-gdbserver"
I get:
target remote localhost:2347^M
Remote debugging using localhost:2347^M
Reading symbols from /lib64/ld-linux-x86-64.so.2...^M
Reading symbols from /usr/lib/debug//lib/x86_64-linux-gnu/ld-2.31.so...^M
0x00007ffff7fd0100 in _start () from /lib64/ld-linux-x86-64.so.2^M
(gdb) continue^M
Continuing.^M
Cannot execute this command while the target is running.^M
Use the "interrupt" command to stop the target^M
and then try again.^M
(gdb) FAIL: gdb.multi/multi-re-run.exp: re_run_inf=1: iter=1: runto: run to all_started
The test does:
- Connect to a remote target with inferior 2, continue and stop on the
all_started function
- Connect to a separate remote target / GDBserver instance with inferior 1,
continue and (expect to) stop on the all_started function
The failure seen above happens when trying to continue inferior 1.
What happens is:
- GDB tells inferior 1's remote target to continue
- We go into fetch_inferior_event, try to get an event at random from
the targets
- do_target_wait happens to pick inferior 2's target
- That target return TARGET_WAITKIND_NO_RESUMED, which makes sense:
inferior 2 is stopped, that target has no resumed thread
- handle_no_resumed tries to update the thread list of all targets:
for (auto *target : all_non_exited_process_targets ())
{
switch_to_target_no_thread (target);
update_thread_list ();
}
- When trying to update the thread list of inferior 1's target, it hits
the "Cannot execute this command while the target is running" error.
This target is working in "remote all-stop" mode, and it is currently
waiting for a stop reply, so it can't send packets to update the
thread list at this time.
To handle the problem described in the comment in handle_no_resumed, I
don't think it is necessary to update the thread list of all targets,
but only the event target. That comment describes a kind of race
condition where some target reports a breakpoint hit for a thread and
then its last remaining resumed thread exits, so sends a "no resumed"
event. If we ended up resuming the thread that hit a breakpoint, we
want to ignore the "no resumed" and carry on.
But I don't really see why we need to update the thread list on the
other targets. I can't really articulate this, it's more a gut feeling,
maybe I just fail to imagine the situation where this is needed. But
here is the patch anyway, so we can discuss it. The patch changes
handle_no_resumed to only update the thread list of the event target.
This fixes the test run shown above.
The way I originally tried to fix this was to make
remote_target::update_thread_list return early if the target is
currently awaiting a stop reply, since there's no way it can update the
thread list at that time. But that felt more like papering over the
problem. I then thought that we probably shouldn't be asking the target
to update the thread list unnecessarily.
Change-Id: Ide3df22b4f556478e155ad1c67ad4b4fe7c26a58
Running
$ ../gdbserver/gdbserver --once --attach :1234 539436
with ASan while /proc/sys/kernel/yama/ptrace_scope is set to 1 (prevents
attaching) shows that we fail to free some platform-specific objects
tied to the process_info (process_info_private and arch_process_info):
Direct leak of 32 byte(s) in 1 object(s) allocated from:
#0 0x7f6b558b3fb9 in __interceptor_calloc /usr/src/debug/gcc/libsanitizer/asan/asan_malloc_linux.cpp:154
#1 0x562eaf15d04a in xcalloc /home/simark/src/binutils-gdb/gdbserver/../gdb/alloc.c:100
#2 0x562eaf251548 in xcnew<process_info_private> /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/poison.h:122
#3 0x562eaf22810c in linux_process_target::add_linux_process_no_mem_file(int, int) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:426
#4 0x562eaf22d33f in linux_process_target::attach(unsigned long) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:1132
#5 0x562eaf1a7222 in attach_inferior /home/simark/src/binutils-gdb/gdbserver/server.cc:308
#6 0x562eaf1c1016 in captured_main /home/simark/src/binutils-gdb/gdbserver/server.cc:3949
#7 0x562eaf1c1d60 in main /home/simark/src/binutils-gdb/gdbserver/server.cc:4084
#8 0x7f6b552f630f in __libc_start_call_main (/usr/lib/libc.so.6+0x2d30f)
Indirect leak of 56 byte(s) in 1 object(s) allocated from:
#0 0x7f6b558b3fb9 in __interceptor_calloc /usr/src/debug/gcc/libsanitizer/asan/asan_malloc_linux.cpp:154
#1 0x562eaf15d04a in xcalloc /home/simark/src/binutils-gdb/gdbserver/../gdb/alloc.c:100
#2 0x562eaf2a0d79 in xcnew<arch_process_info> /home/simark/src/binutils-gdb/gdbserver/../gdbsupport/poison.h:122
#3 0x562eaf295e2c in x86_target::low_new_process() /home/simark/src/binutils-gdb/gdbserver/linux-x86-low.cc:723
#4 0x562eaf22819b in linux_process_target::add_linux_process_no_mem_file(int, int) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:428
#5 0x562eaf22d33f in linux_process_target::attach(unsigned long) /home/simark/src/binutils-gdb/gdbserver/linux-low.cc:1132
#6 0x562eaf1a7222 in attach_inferior /home/simark/src/binutils-gdb/gdbserver/server.cc:308
#7 0x562eaf1c1016 in captured_main /home/simark/src/binutils-gdb/gdbserver/server.cc:3949
#8 0x562eaf1c1d60 in main /home/simark/src/binutils-gdb/gdbserver/server.cc:4084
#9 0x7f6b552f630f in __libc_start_call_main (/usr/lib/libc.so.6+0x2d30f)
Those objects are deleted by linux_process_target::mourn, but that is
not called if we fail to attach, we only call remove_process. I
initially fixed this by making linux_process_target::attach call
linux_process_target::mourn on failure (before calling error). But this
isn't done anywhere else (including in GDB) so it would just be
confusing to do things differently here.
Instead, add a linux_process_target::remove_linux_process helper method
(which calls remove_process), and call that instead of remove_process in
the Linux target. Move the free-ing of the extra data from the mourn
method to that new method.
Change-Id: I277059a69d5f08087a7f3ef0b8f1792a1fcf7a85
This commit replaces an earlier commit that worked around the issues
reported in bug PR gdb/28833.
The previous commit just implemented a work around in order to avoid
the worst results of the bug, but was not a complete solution. The
full solution was considered too risky to merge close to branching GDB
12. This improved fix has been applied after GDB 12 branched. See
this thread for more details:
https://sourceware.org/pipermail/gdb-patches/2022-March/186391.html
This commit replaces this earlier commit:
commit 74a159a420d4b466cc81061c16d444568e36740c
Date: Fri Mar 11 14:44:03 2022 +0000
gdb: work around prompt corruption caused by bracketed-paste-mode
Please read that commit for a full description of the bug, and why is
occurs.
In this commit I extend GDB to use readline's rl_deprep_term_function
hook to call a new function gdb_rl_deprep_term_function. From this
new function we can now print the 'quit' message, this replaces the
old printing of 'quit' in command_line_handler. Of course, we only
print 'quit' in gdb_rl_deprep_term_function when we are handling EOF,
but thanks to the previous commit (to readline) we now know when this
is.
There are two aspects of this commit that are worth further
discussion, the first is in the new gdb_rl_deprep_term_function
function. In here I have used a scoped_restore_tmpl to disable the
readline global variable rl_eof_found.
The reason for this is that, in rl_deprep_terminal, readline will
print an extra '\n' character before printing the escape sequence to
leave bracketed paste mode. You might then think that in the
gdb_rl_deprep_term_function function, we could simply print "quit" and
rely on rl_deprep_terminal to print the trailing '\n'. However,
rl_deprep_terminal only prints the '\n' when bracketed paste mode is
on. If the user has turned this feature off, no '\n' is printed.
This means that in gdb_rl_deprep_term_function we need to print
"quit" when bracketed paste mode is on, and "quit\n" when bracketed
paste mode is off.
We could absolutely do that, no problem, but given we know how
rl_deprep_terminal is implemented, it's easier (I think) to just
temporarily clear rl_eof_found, this prevents the '\n' being printed
from rl_deprep_terminal, and so in gdb_rl_deprep_term_function, we can
now always print "quit\n" and this works for all cases.
The second issue that should be discussed is backwards compatibility
with older versions of readline. GDB can be built against the system
readline, which might be older than the version contained within GDB's
tree. If this is the case then the system readline might not contain
the fixes needed to support correctly printing the 'quit' string.
To handle this situation I have retained the existing code in
command_line_handler for printing 'quit', however, this code is only
used now if the version of readline we are using doesn't not include
the required fixes. And so, if a user is using an older version of
readline, and they have bracketed paste mode on, then they will see
the 'quit' sting printed on the line below the prompt, like this:
(gdb)
quit
I think this is the best we can do when someone builds GDB against an
older version of readline.
Using a newer version of readline, or the patched version of readline
that is in-tree, will now give a result like this in all cases:
(gdb) quit
Which is what we want.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
This commit is a partial back-port of this upstream readline commit:
commit 002d31aa5f5929eb32d0e0e2e8b8d35d99e59961
Author: Chet Ramey <chet.ramey@case.edu>
Date: Thu Mar 3 11:11:47 2022 -0500
add rl_eof_found to public API; fix pointer aliasing problems \
with history-search-backward; fix a display problem with \
runs of invisible characters at the end of a physical \
screen line
I have only pulled in the parts of this commit that relate to the new
rl_eof_found global, and the RL_STATE_EOF state flag. These changes
are needed in order to fix PR cli/28833, and are discussed in this
thread to the bug-readline mailing list:
https://lists.gnu.org/archive/html/bug-readline/2022-02/msg00021.html
The above commit is not yet in any official readline release, but my
hope is that now it has been merged into the readline tree it should
be safe enough to back port this fix to GDB's tree.
At some point in the future we will inevitably want to roll forward
the version of readline that we maintain in the binutils-gdb
repository. When that day comes the changes in this commit can be
replaced with the latest upstream readline code, as I have not changed
the meaning of this code at all from what is in upstream readline.
This commit alone does not fix the PR cli/28833 issue, for that see
the next commit, which changes GDB itself.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
In this commit:
commit a6b413d24c
Date: Fri Mar 11 14:44:03 2022 +0000
gdb: work around prompt corruption caused by bracketed-paste-mode
a change was made to GDB to work around bug PR gdb/28833. The
consequence of this work around is that, when bracketed paste mode is
enabled in readline, and GDB is quit by sending EOF, then the output
will look like this:
(gdb)
quit
The ideal output, which is what we get when bracketed paste mode is
off, is this:
(gdb) quit
The reason we need to make this change is explained in the original
commit referenced above. What isn't mentioned in the above commit, is
that the change that motivated this work around was only added in
readline 8, older versions of readline don't require the change.
In later commits in this series I will add a fix to GDB's in-tree copy
of readline (this fix is back-ported from upstream readline), and then
I will change GDB so that, when using the (patched) in-tree readline,
we can have the ideal output in all cases.
However, GDB can be built against the system readline. When this is
done, and the system readline is version 8, then we will still have to
use the work around (two line) style output.
But, if GDB is built against the system readline, and the system
readline is an older version 7 readline, then there's no reason why we
can't have the ideal output, after all, readline 7 doesn't include the
change that we need to work around.
This commit changes GDB so that, when using readline 7 we get the
ideal output in all cases. This change is trivial (a simple check
against the readline version number) so I think this should be fine to
include.
For testing this commit, you need to configure GDB including the
'--with-system-readline' flag, and build GDB on a system that uses
readline 7, for example 'Ubuntu 18.04'. Then run the test
'gdb.base/eof-exit.exp', you should expect everything to PASS.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28833
This test sometimes fail like this:
info threads^M
Id Target Id Frame ^M
11.12 process 2270719 Couldn't get registers: No such process.^M
(gdb) FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: no threads left
[Inferior 11 (process 2270719) exited normally]^M
info inferiors^M
Num Description Connection Executable ^M
* 1 <null> /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/fork-plus-threads/fork-plus-threads ^M
11 <null> /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.threads/fork-plus-threads/fork-plus-threads ^M
(gdb) FAIL: gdb.threads/fork-plus-threads.exp: detach-on-fork=off: only inferior 1 left (the program exited)
I can get it to fail quite reliably by pinning it to a core:
$ taskset -c 5 make check TESTS="gdb.threads/fork-plus-threads.exp"
The previous attempt at fixing this was:
https://sourceware.org/pipermail/gdb-patches/2021-October/182846.html
What we see is part due to a possible unfortunate ordering of events
given by the kernel, and what could be considered a bug in GDB.
The test program makes a number of forks, waits them all, then exits.
Most of the time, GDB will get and process the exit event for inferior 1
after the exit events of all the children. But this is not guaranteed.
After the last child exits and is waited by the parent, the parent can
exit quickly, such that GDB collects from the kernel the exit events for
the parent and that child at the same time. It then chooses one event
at random, which can be the event for the parent. This will result in
the parent appearing to exit before its child. There's not much we can
do about it, so I think we have to adjust the test to cope.
After expect has seen the "exited normally" notification for inferior 1,
it immediately does an "info thread" that it expects to come back empty.
But at this point, GDB might not have processed inferior 11's (the last
child) exit event, so it will look like there is still a thread. Of
course that thread is dead, we just don't know it yet. But that makes
the "no thread" test fail. If the test waited just a bit more for the
"exited normally" notification for inferior 11, then the list of threads
would be empty.
So, first change, make the test collect all the "exited normally"
notifications for all inferiors before proceeding, that should ensure we
see an empty thread list. That would fix the first FAIL above.
However, we would still have the second FAIL, as we expect inferior 11
to not be there, it should have been deleted automatically. Inferior 11
is normally deleted when prune_inferiors is called. That is called by
normal_stop, which is only called by fetch_inferior_event only if the
event thread completed an execution command FSM (thread_fsm). But the
FSM for the continue command completed when inferior 1 exited. At that
point inferior 11 was not prunable, as it still had a thread. When
inferior 11 exits, prune_inferiors is not called.
I think that can be considered a GDB bug. From the user point of view,
there's no reason why in one case inferior 11 would be deleted and not
in the other case.
This patch makes the somewhat naive change to call prune_inferiors in
fetch_inferior_event, so that it is called in this case. It is placed
at this particular point in the function so that it is called after the
user inferior / thread selection is restored. If it was called before
that, inferior 11 wouldn't be pruned, because it would still be the
current inferior.
Change-Id: I48a15d118f30b1c72c528a9f805ed4974170484a
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=26272
This fixes a build breakage in my recent coff-pe-read.c change.
I'm sorry about this. I don't understand how it happened, because I
definitely built and tested the series on Windows, and I didn't change
it before pushing. Something must have gone wrong on the Windows
build, but I don't know what. I'll investigate and and re-test to be
sure.
coff-pe-read.c uses xsnprintf and alloca, but using std::string is
better, and just as easy. In general I think alloca is something to
be avoided, and unbounded uses especially so.
