Simon Marchi [Sun, 24 Dec 2023 15:38:35 +0000 (10:38 -0500)]
gdb: remove VALUE_REGNUM, add value::regnum
Remove VALUE_REGNUM, replace it with a method on struct value. Set
`m_location.reg.regnum` directly from value::allocate_register_lazy,
which is fine because allocate_register_lazy is a static creation
function for struct value.
Remove VALUE_NEXT_FRAME_ID, replace it with a method on struct value. Set
`m_location.reg.next_frame_id` directly from value::allocate_register_lazy,
which is fine because allocate_register_lazy is a static creation
function for struct value.
Simon Marchi [Thu, 21 Dec 2023 16:32:55 +0000 (16:32 +0000)]
gdb: implement address_from_register using value_from_register
As explained in the comment removed by the previous commit "gdb: pass
non-nullptr frame to gdbarch_value_from_register in
address_from_register", address_from_register copies some implementation
bits from value_from_register:
/* This routine may be called during early unwinding, at a time
where the ID of FRAME is not yet known. Calling value_from_register
would therefore abort in get_frame_id. However, since we only need
a temporary value that is never used as lvalue, we actually do not
really need to set its VALUE_NEXT_FRAME_ID. Therefore, we re-implement
the core of value_from_register, but use the null_frame_id. */
This is no longer relevant, since we now create a value with a valid next
frame id, so change address_from_register to use value_from_register.
By now, all register struct values should have a valid next frame id
(assuming they are created using value::allocate_register or
value::allocate_register_lazy), so there should be no need to pass a
frame alongside the value to read_frame_register_value. Remove the
frame parameter and adjust read_frame_register_value accordingly.
While at it, make read_frame_register_value static, it's only used in
findvar.c.
Simon Marchi [Thu, 21 Dec 2023 16:21:52 +0000 (16:21 +0000)]
gdb: add type parameter to value::allocate_register and add value::allocate_register_lazy
Some places that create register struct values don't use register_type
to obtain the value type. This prevents them from using the current
version of value::allocate_register. One spot (value_of_register_lazy)
also creates a lazy register value.
Add a value::allocate_register_lazy method. Add some type parameters
to value::allocate_register and value::allocate_register_lazy, to let
the caller specify the type to use for the value. The parameters
default to nullptr, in which case we use register_type to obtain the
type.
Simon Marchi [Thu, 21 Dec 2023 16:10:17 +0000 (16:10 +0000)]
gdb: pass non-nullptr frame to gdbarch_value_from_register in address_from_register
address_from_register used to pass null_frame_id to
gdbarch_value_from_register as "this frame"'s id, because it's possible
for it to be called during unwind, when "this frame"'s id is not yet
known. This create an oddity where those register struct values are
created without a valid next frame id. I would much prefer for things
to be consistent and have all register struct values to have a valid
next frame id.
Since gdbarch_value_from_register takes a frame_info_ptr now, rather
than a frame_id, we can pass down "this frame", even if it doesn't have
a valid id. gdbarch_value_from_register implementations can obtain the
next frame from it.
However, it's possible for the "this frame"'s next frame to be an
inline frame, inlined in "this frame", in which case that next frame's
id is also not known. So, loop until we get to the next non-inline
frame (which is actually the frame where registers for "this frame" are
unwound from). This is the same thing that we do in
value_of_register_lazy, for the same reason. A later patch will factor
out this "while next frame is inline" loop to apply it to all register
struct values, so this is somewhat temporary.
Simon Marchi [Thu, 21 Dec 2023 16:03:02 +0000 (16:03 +0000)]
gdb: pass frame_info_ptr to gdbarch_value_from_register
Pass a frame_info_ptr rather than a frame_id. This avoids having to do
a frame lookup on the callee side, when we can just pass the frame down
directly.
I think this fixes a bug in rs6000-tdep.c where the id of the wrong
frame was set to `VALUE_NEXT_FRAME_ID (v)`.
Mike Frysinger [Sun, 24 Dec 2023 10:06:57 +0000 (05:06 -0500)]
sim: cgen: mark cgen_rtx_error noreturn
Since this function never returns, mark it as such to fix some unused
variable warnings in error code paths.
For example, cris triggers:
sim/cris/semcrisv10f-switch.c:3558:11: error:
variable 'tmp_newval' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
Even though it has an "else" path that calls this error function.
Mike Frysinger [Sun, 24 Dec 2023 08:59:02 +0000 (03:59 -0500)]
sim: sh: refine pwsb & pwad nops
Since these insns don't do anything and are effectively ignored,
return early to avoid doing any common processing at the end as
that requires initializing variables like "res" with something.
Mike Frysinger [Sun, 24 Dec 2023 08:53:03 +0000 (03:53 -0500)]
sim: sh: fix plds Dz,MACL implementation
The plds Dz,MACL insn stores the Dz bit into MACL. The current code
was storing the "res" variable into Dz and then into MACL, but not
setting "res" to anything. Delete that logic and make it match the
existing plds Dz,MACH insn.
Mike Frysinger [Fri, 22 Dec 2023 23:05:50 +0000 (18:05 -0500)]
sim: m32c: fix -Wshadow=local warnings
These decoders declare a lot of common variables for use by substeps,
and then shadows a few because of how the opc generator is implemented.
Easiest way around it is to rename the per-substep vars as needed as
anything more would require substantial changes to the opc logic.
Mike Frysinger [Sat, 23 Dec 2023 04:17:45 +0000 (23:17 -0500)]
sim: cris: disable -Wshadow=local in generated mloop files
The mloop files include CGEN generated switch files which have some
nested assignments that expand into repeated shadowed variables.
Fixing this looks fairly non-trivial as it appears to be interplay
between the common CGEN code and how this particular set of cris
insns are defined. Disable the warning instead.
In file included from sim/cris/mloop.in:286:
sim/cris/semcrisv10f-switch.c: In function ‘crisv10f_engine_run_full’:
sim/cris/semcrisv10f-switch.c:12383:8: error: declaration of ‘opval’ shadows a previous local [-Werror=shadow=local]
12383 | SI opval = tmp_addr;
| ^~~~~
sim/cris/semcrisv10f-switch.c:12371:9: note: shadowed declaration is here
12371 | USI opval = ({ SI tmp_addr;
| ^~~~~
And the code looks like:
USI opval = ({
...
{
SI opval = tmp_addr;
...
}
...
});
Since the CGEN code treats "opval" as an internal variable that the cpu
definitions don't have direct access to, the likelihood of this being a
real bug is low, so leave it be. The warning is suppressed for more code
that is hand written (e.g. the mloop logic), but disabling for the entire
file is the easiest way to suppress while keeping it on everywhere else in
the sim.
Mike Frysinger [Fri, 22 Dec 2023 01:03:41 +0000 (20:03 -0500)]
sim: arm: fix -Wshadow=local warnings
Remove duplicate nested variable declarations, rename some to avoid
confusion when the type is different or the original value should be
retained, and fix some weirdness with nested enums in structs.
Mike Frysinger [Fri, 22 Dec 2023 01:01:31 +0000 (20:01 -0500)]
sim: aarch64: fix -Wshadow=local warnings
These functions have local vars named "val" of type float, and
then create nested vars named "val" of type double. This is a
bit confusing and causes build time warnings.
Tom Tromey [Tue, 12 Dec 2023 17:09:55 +0000 (10:09 -0700)]
Avoid exception from attach in DAP
I noticed that the DAP attach test case (and similarly
remoted-dap.exp) had a rogue exception stack trace in the log. It
turns out that an attach will generate a stop that does not have a
reason.
This patch fixes the problem in the _on_stop event listener by making
it a bit more careful when examining the event reason. It also adds
some machinery so that attach stops can be suppressed, which I think
is the right thing to do.
Reviewed-By: Kévin Le Gouguec <legouguec@adacore.com>
Tom Tromey [Tue, 12 Dec 2023 16:30:41 +0000 (09:30 -0700)]
Add DAP log level parameter
This adds a new parameter to control the DAP logging level. By
default, "expected" exceptions are not logged, but the parameter lets
the user change this when more logging is desired.
This also changes a couple of spots to avoid logging the stack trace
for a DAPException.
This patch also documents the existing DAP logging parameter. I
forgot to document this before.
Reviewed-By: Eli Zaretskii <eliz@gnu.org> Reviewed-By: Kévin Le Gouguec <legouguec@adacore.com>
Tom Tromey [Tue, 12 Dec 2023 16:29:43 +0000 (09:29 -0700)]
Introduce and use DAPException
This introduces a new DAPException class, and then changes various
spots in the DAP implementation to wrap "expected" exceptions in this.
This class will help detect rogue exceptions caused by bugs in the
implementation.
Reviewed-By: Kévin Le Gouguec <legouguec@adacore.com>
Tom Tromey [Mon, 11 Dec 2023 17:04:23 +0000 (10:04 -0700)]
Fix build with clang 16
clang 16 reports a missing declaration in new-op.cc. We believed
these operators to be declared starting with C++14, but apparently
that is not the case.
This patch reverts the earlier change and then updates the comment to
reflect the current state.
Kévin Le Gouguec [Fri, 22 Dec 2023 13:06:15 +0000 (14:06 +0100)]
gdb: fix refactoring hiccup in rs6000_register_to_value
In 2023-12-14 "gdb: make get_frame_register_bytes take the next frame"
(9fc79b42369), *_register_to_value functions were made to (a) call
get_next_frame_sentinel_okay (frame) (b) pass that next frame to
get_frame_register_bytes.
Step (b) was omitted for rs6000-tdep.c; this manifests as a regression on
PPC platforms for e.g. O2_float_param: instead of seeing…
Temporary breakpoint 1, callee.increment (val=val@entry=99.0, msg=...) at callee.adb:19
… we get "optimized_out" for val. Passing next_frame to
get_frame_register_bytes fixes the issue.
Approved-By: Simon Marchi <simon.marchi@efficios.com>
Tom Tromey [Thu, 7 Dec 2023 16:51:52 +0000 (09:51 -0700)]
Add 'program' to DAP 'attach' request
In many cases, it's not possible for gdb to discover the executable
when a DAP 'attach' request is used. This patch lets the IDE supply
this information.
Guinevere Larsen [Tue, 16 May 2023 14:25:53 +0000 (16:25 +0200)]
gdb: add git trailer information on gdb/MAINTAINERS
The project has been using Tested-By (tb), Reviewed-By (rb) and
Approved-By (ab) for some time, but there has been no information to be
found in the actual repository. This commit changes that by adding
information about all git trailers to the MAINTAINERS file, so that it
can be easily double-checked. Simply put, the trailers in use work as
follows:
* Tested-by: The person tested the patch and it fixes the problem, or
introduces no regressions (or both).
* Acked-by: The general outline looks good, but the maintainer hasn't
looked at the code
* Reviewed-by: The code looks good, but the reviewer has not approved
the patch to go upstream
* Approved-by: The patch is ready to be pushed to master
These last 3 trailers can also be restricted to one or more areas of GDB
by adding the areas in a comma separated list in parenthesis after the
trailers.
Finally, for completeness sake, the trailers Co-Authored-By and Bug
were added, even though they have been in use for a long time already
Reviewed-By: Kevin Buettner <kevinb@redhat.com> Reviewed-By: Luis Machado <luis.machado@arm.com> Approved-By: John Baldwin <jhb@FreeBSD.org>
Jan Beulich [Fri, 22 Dec 2023 08:35:52 +0000 (09:35 +0100)]
hppa/ELF: fix .text/.data interaction with .previous
For some ELF targets .text/.data are overridden. In that case
obj_elf_{text,data}() need calling, just like .code vectors to that
function for the remaining ELF targets.
While there also hand on the function arguments, even if right now
they're meaningless. This matches what other targets' code does.
Jan Beulich [Fri, 22 Dec 2023 08:35:02 +0000 (09:35 +0100)]
RISC-V: drop .bss override
It doesn't look to be a good idea to override the custom handler that
ELF has; afaict doing so broke .previous, and a sub-section specifier
wasn't accepted either.
Jan Beulich [Fri, 22 Dec 2023 08:34:10 +0000 (09:34 +0100)]
x86-64: refuse "high" 8-bit regs with .insn and VEX/XOP/EVEX encodings
Much like REX, those encodings - if permitting 8-bit regs at all, i.e.
only starting with APX - permit use of "new" 8-bit registers only. %ah,
%ch, %dh, and %bh cannot be encoded and hence should be rejected.
Permit their use outside of 64-bit code though, as "new" registers
simply don't exist there.
Jan Beulich [Fri, 22 Dec 2023 08:33:12 +0000 (09:33 +0100)]
x86: properly respect rex/{rex}
This addresses two issues: For one, a user specified "rex" did not cause
the diagnostic to trigger when there was no other need for a REX prefix;
instead, another than the specified insn+operands was encoded. And then
(which is what this started from) .insn didn't respect {rex} (and was
otherwise similarly flawed as ordinary insns).
The latter requires splitting out code from md_assemble(), for it to
become re-usable. Besides the addition to address the first issue, that
code then also needs generalizing to account for immediate operands, as
with .insn we can't make assumptions anymore on all respective templates
having at most two operands (we still can build upon there being at most
two non-immediate operands, though). While moving the code also simplify
the first if(), by folding redundant checks.
In the new testcase also test a few more things which afaics weren't
tested till now.
Mike Frysinger [Wed, 20 Dec 2023 01:04:34 +0000 (20:04 -0500)]
sim: common: add $LINENO rewriting support to genmloop scripts
The generated mloop files can trigger compile time warnings. It can
be difficult to see/understand where the original code is coming from
as all the diagnostics point to the generated output. Using #line
pragmas, we can point people to the original source files.
Unfortunately, this code is written in POSIX shell, and that lacks
support for line number tracking. The $LINENO variable, even when
available, can just be plain wrong. For example, when using dash
and subshells, $LINENO can end up having negative values. Add a
wrapper script that will uses awk to rewrite the $LINENO variable
to the right value to avoid all that.
Basically lineno.sh takes an input script, rewrites all uses of
$LINENO into the actual line number (and $0 into the original file
name), and then executes the temporary script.
This commit doesn't actually add #line pragmas to any files. That
comes next.
Tom Tromey [Fri, 8 Dec 2023 17:32:07 +0000 (10:32 -0700)]
Rename TUI locator window -> status
The TUI status window is called the "locator" in the source, but
"status" in the documentation. Whenever I've needed to find the code,
I've had to search to "locate" it (ha, ha). This patch renames the
window to match the public name of the window.
Tom Tromey [Fri, 8 Dec 2023 17:14:41 +0000 (10:14 -0700)]
Rename tui-stack -> tui-status
The TUI status line is called the "status" window in the
documentation, but not in the source. There, the relevant files are
named "tui-stack", which to me makes it sound like they have something
to do with backtraces. This patch renames them to "tui-status".
Rainer Orth [Thu, 21 Dec 2023 11:51:26 +0000 (12:51 +0100)]
ld: Add lib32 directories for 32-bit emulation on FreeBSD/amd64
GNU ld currently fails to link 32-bit executables on FreeBSD/amd64 when
the linked libraries have dependencies on shared objects themselves:
$ gcc -m32 -o ei ei.c -lexecinfo
/var/gcc/binutils/amd64/lib/gcc/amd64-pc-freebsd14.0/13.2.0/../../../../amd64-pc-freebsd14.0/bin/ld:
warning: libelf.so.2, needed by /usr/lib/../lib32/libexecinfo.so, not found
(try using -rpath or -rpath-link)
/var/gcc/binutils/amd64/lib/gcc/amd64-pc-freebsd14.0/13.2.0/../../../../amd64-pc-freebsd14.0/bin/ld:
/usr/lib/../lib32/libexecinfo.so: undefined reference to `elf_begin@R1.0'
[...]
Pedro Alves [Thu, 21 Dec 2023 10:43:20 +0000 (10:43 +0000)]
Fix Clang build issue with flexible array member and non-trivial dtor
Commit d5cebea18e7a ("Make cached_reg_t own its data") added a
destructor to cached_reg_t.
That caused a build problem with Clang, which errors out like so:
> CXX python/py-unwind.o
> gdb/python/py-unwind.c:126:16: error: flexible array member 'reg' of type 'cached_reg_t[]' with non-trivial destruction
> 126 | cached_reg_t reg[];
> | ^
This is is not really a problem for our code, which allocates the
whole structure with xmalloc, and then initializes the array elements
with in-place new, and then takes care to call the destructor
manually. Like, commit d5cebea18e7a did:
for (int i = 0; i < cached_frame->reg_count; i++)
- xfree (cached_frame->reg[i].data);
+ cached_frame->reg[i].~cached_reg_t ();
Maybe we should get rid of the flexible array member and use a bog
standard std::vector. I doubt this would cause any visible
performance issue.
Meanwhile, to unbreak the build, this commit switches from C99-style
flexible array member to 0-length array. It behaves the same, and
Clang doesn't complain. I got the idea from here:
Mike Frysinger [Thu, 21 Dec 2023 06:25:50 +0000 (01:25 -0500)]
sim: aarch64: fix -Wimplicit-fallthrough warnings
Replace some fall through comments with the attribute, and add some
default abort calls when the compiler can't figure out that the set
of values were already fully enumerated in the switch statement.
Mike Frysinger [Thu, 21 Dec 2023 05:38:31 +0000 (00:38 -0500)]
sim: signal: mark signal callback funcs as noreturn since they don't return
All funcs already call other funcs that don't return. The mips port is
the only exception because its generic exception handler can return in
the case of normal exceptions. So while the exceptions its signal handler
triggers doesn't return, we can't express that conditional logic. So add
some useless abort calls to make the compiler happy.
Mike Frysinger [Thu, 21 Dec 2023 05:00:49 +0000 (00:00 -0500)]
sim: mn10300: fix LAST_TIMER_REG typo
The compiler pointed out that we're testing LAST_TIMER_REG and
LAST_COUNTER which are the same value ... and that's because we
set LAST_TIMER_REG to the wrong register. Fix the typo.
Mike Frysinger [Thu, 21 Dec 2023 04:59:28 +0000 (23:59 -0500)]
sim: bfin: clean up astat reg name decode a little
The compiler pointed out we checked AZ twice. Sort by name to avoid
that in the future, and to make it clearer that we have coverage of
all the bits. And add the bits we were missing.
The order here doesn't matter as it's just turning a pointer into a
human readable string when store tracing is enabled.
Mike Frysinger [Tue, 19 Dec 2023 10:47:32 +0000 (05:47 -0500)]
sim: common: delete unused scache in some mloop paths
The scache vars aren't used by ports in the pbb & fast codepaths,
nor are they documented as inputs to the callbacks, so delete them
to avoid unused variable compiler warnings.
Mike Frysinger [Wed, 20 Dec 2023 00:54:13 +0000 (19:54 -0500)]
sim: cgen: unify the genmloop logic a bit
Pull out the common parts of the genmloop invocation into the common
code. This will make it easier to add more, and make the per-port
differences a little more obvious.
gprofng: 31169 Source code locations can not be found in a C++ application
gprofng incorrectly reads the form of the DW_FORM_ref_addr attribute for DWARF
Version 3 or later.
From DWARF specification:
References that use the attribute form DW_FORM_ref_addr are specified to
be four bytes in the DWARF 32-bit format and eight bytes in the DWARF
64-bit format, while DWARF Version 2 specifies that such references have
the same size as an address on the target system.
2023-12-18 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
PR gprofng/31169
* src/DwarfLib.cc: Fix the reader for DW_FORM_ref_addr.
Pedro Alves [Fri, 1 Dec 2023 17:45:21 +0000 (17:45 +0000)]
Fix handling of vanishing threads that were stepping/stopping
Downstream, AMD is carrying a testcase
(gdb.rocm/continue-over-kernel-exit.exp) that exposes a couple issues
with the amd-dbgapi target's handling of exited threads. The test
can't be added upstream yet, unfortunately, due to dependency on DWARF
extensions that can't be upstreamed yet. However, it can be found on
the mailing list on the same series as this patch.
The test spawns a kernel with a number of waves. The waves do nothing
but exit. There is a breakpoint on the s_endpgm instruction. Once
that breakpoint is hit, the test issues a "continue" command. We
should see one breakpoint hit per wave, and then the whole program
exiting. We do see that, however we also see this:
That "New AMDGPU Wave" output comes from infrun.c itself adding the
thread to the GDB thread list, because it got an event for a thread
not on the thread list yet. The output shows "?"s instead of proper
coordinates, because the event was a TARGET_WAITKIND_THREAD_EXITED,
i.e., the wave was already gone when infrun.c added the thread to the
thread list.
That shouldn't ever happen for the amd-dbgapi target, threads should
only ever be added by the backend.
Note "New AMDGPU Wave ?:?:?:1" is for wave 1. What happened was that
wave 1 terminated previously, and a previous call to
amd_dbgapi_target::update_thread_list() noticed the wave had vanished
and removed it from the GDB thread list. However, because the wave
was stepping when it terminated (due to the displaced step over the
s_endpgm) instruction, it is guaranteed that the amd-dbgapi library
queues a WAVE_COMMAND_TERMINATED event for the exit.
When we process that WAVE_COMMAND_TERMINATED event, in
amd-dbgapi-target.c:process_one_event, we return it to the core as a
TARGET_WAITKIND_THREAD_EXITED event:
Recall the wave is already gone from the GDB thread list. So when GDB
sees that TARGET_WAITKIND_THREAD_EXITED event for a thread it doesn't
know about, it adds the thread to the thread list, resulting in that:
[New AMDGPU Wave ?:?:?:1 (?,?,?)/?]
and then, because it was a TARGET_WAITKIND_THREAD_EXITED event, GDB
marks the thread exited right afterwards:
[AMDGPU Wave ?:?:?:1 (?,?,?)/? exited]
The fix is to make amd_dbgapi_target::update_thread_list() _not_
delete vanishing waves iff they were stepping or in progress of being
stopped. These two cases are the ones dbgapi guarantees will result
in a WAVE_COMMAND_TERMINATED event if the wave terminates:
/**
* A command for a wave was not able to complete because the wave has
* terminated.
*
* Commands that can result in this event are ::amd_dbgapi_wave_stop and
* ::amd_dbgapi_wave_resume in single step mode. Since the wave terminated
* before stopping, this event will be reported instead of
* ::AMD_DBGAPI_EVENT_KIND_WAVE_STOP.
*
* The wave that terminated is available by the ::AMD_DBGAPI_EVENT_INFO_WAVE
* query. However, the wave will be invalid since it has already terminated.
* It is the client's responsibility to know what command was being performed
* and was unable to complete due to the wave terminating.
*/
AMD_DBGAPI_EVENT_KIND_WAVE_COMMAND_TERMINATED = 2,
As the comment says, it's GDB's responsability to know whether the
wave was stepping or being stopped. Since we now have a wave_info map
with one entry for each wave, that seems like the place to store that
information. However, I still decided to put all the coordinate
information in its own structure. I.e., basically renamed the
existing wave_info to wave_coordinates, and then added a new wave_info
structure that holds the new state, plus a wave_coordinates object.
This seemed cleaner as there are places where we only need to
instantiate a wave_coordinates object.
There's an extra twist. The testcase also exercises stopping at a new
kernel right after the first kernel fully exits. In that scenario, we
were hitting this assertion after the first kernel fully exits and the
hit of the breakpoint at the second kernel is handled:
[amd-dbgapi] process_event_queue: Pulled event from dbgapi: event_id.handle = 26, event_kind = WAVE_STOP
[amd-dbgapi-lib] suspending queue_3, queue_2, queue_1 (refresh wave list)
../../src/gdb/amd-dbgapi-target.c:1625: internal-error: amd_dbgapi_thread_deleted: Assertion `it != info->wave_info_map.end ()' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
This is the exact same problem as above, just a different
manifestation. In this scenario, we end up in update_thread_list
successfully deleting the exited thread (because it was no longer the
current thread) that was incorrectly added by infrun.c. Because it
was added by infrun.c and not by amd-dbgapi-target.c:add_gpu_thread,
it doesn't have an entry in the wave_info map, so
amd_dbgapi_thread_deleted trips on this assertion:
The testcase thus tries both running to exit after the first kernel
exits, and running to a breakpoint in a second kernel after the first
kernel exits.
Approved-By: Lancelot Six <lancelot.six@amd.com> (amdgpu)
Change-Id: I43a66f060c35aad1fe0d9ff022ce2afd0537f028
Those '?' are because the thread/wave is already gone by the time GDB
prints the "exited" notification, we can't ask dbgapi for any info
about the wave anymore.
This commit fixes it by caching the wave's coordinates as soon as GDB
sees the wave for the first time, and making
amd_dbgapi_target::pid_to_str use the cached info.
At first I thought of clearing the wave_info object from a
thread_exited observer. However, that is too soon, resulting in this:
Pedro Alves [Fri, 17 Nov 2023 21:45:06 +0000 (21:45 +0000)]
Step over thread exit, always delete the thread non-silently
With AMD GPU debugging, I noticed that when stepping over a breakpoint
placed on top of the s_endpgm instruction inline (displaced=off), GDB
would behave differently -- it wouldn't print the wave exit. E.g:
In the cases we see the "exited" notification, handle_thread_exit is
what first called delete_thread on the exiting thread, which is
non-silent.
With inline stepping, however, handle_thread_exit ends up in
update_thread_list (via restart_threads) before any delete_thread
call. Thus, amd_dbgapi_target::update_thread_list notices that the
wave is gone and deletes it with delete_thread_silent.
This commit fixes it, by making handle_thread_exited call
set_thread_exited (with the default silent=false) early, which emits
the user-visible notification.
Approved-By: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I22ab3145e18d07c99dace45576307b9f9d5d966f
Pedro Alves [Fri, 1 Dec 2023 13:31:00 +0000 (13:31 +0000)]
displaced_step_finish: Don't fetch the regcache of exited threads
displaced_step_finish can be called with event_status.kind ==
TARGET_WAITKIND_THREAD_EXITED, and in that case it is not possible to
get at the already-exited thread's registers.
This patch moves the get_thread_regcache calls to branches that
actually need it, where we know the thread is still alive.
It also adds an assertion to get_thread_regcache, to help catching
these broken cases sooner.
Approved-By: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I63b5eacb3e02a538fc5087c270d8025adfda88c3