[Bug gdb/31465] New: gdb batch mode crashes when execute "gdb.post_event(lambda: gdb.execute("c"))" in a stop event handler

[Bug gdb/31465] New: gdb batch mode crashes when execute "gdb.post_event(lambda: gdb.execute("c"))" in a stop event handler

[ydeng339 at gatech dot edu]

https://sourceware.org/bugzilla/show_bug.cgi?id=31465

            Bug ID: 31465
           Summary: gdb batch mode crashes when execute
                    "gdb.post_event(lambda: gdb.execute("c"))" in a stop
                    event handler
           Product: gdb
           Version: 15.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gdb
          Assignee: unassigned at sourceware dot org
          Reporter: ydeng339 at gatech dot edu
  Target Milestone: ---

I want to debug my program (with multiple breakpoints) automatically. When
reaching a breakpoint, I want to do some operations and continue the execution. 

Here is the code snippet of my python script:
```
def set_breakpoints():
    # Just add multiple breakpoints
    ...

def on_stop(event):
    # Do some operations
    gdb.post_event(lambda: gdb.execute("c"))

gdb.events.stop.connect(on_stop)

gdb.execute("run")
```

And here is my command: "gdb -x [MY_PYTHON_SCRIPT] --batch --args
[MY_PROGRAM]".

And then, gdb crashes with the following logs:
```
Fatal signal: Segmentation fault
----- Backtrace -----
0x55a013532077 ???
0x55a013634859 ???
0x55a013634a22 ???
0x7eff7742c51f ???
        ./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0
0x7eff780093eb ???
0x7eff78009f1b ???
0x55a013797a2a ???
0x55a0137d01d5 ???
0x7eff7742f494 __run_exit_handlers
        ./stdlib/exit.c:113
0x7eff7742f60f __GI_exit
        ./stdlib/exit.c:143
0x55a01385f3e4 ???
0x55a0136f3554 ???
0x55a0136f404e ???
0x55a01348a15f ???
0x7eff77413d8f __libc_start_call_main
        ../sysdeps/nptl/libc_start_call_main.h:58
0x7eff77413e3f __libc_start_main_impl
        ../csu/libc-start.c:392
0x55a01348fbf4 ???
0xffffffffffffffff ???
---------------------
```

Such crash only appears in batch mode. If I run the program in interactive
mode, the program will not crash (but interactive mode is inefficient). Based
on my test, this bug appears in version 12.1 and version 15.0.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/31465] gdb batch mode crashes when execute "gdb.post_event(lambda: gdb.execute("c"))" in a stop event handler

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31465

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tromey at sourceware dot org

--- Comment #1 from Tom Tromey <tromey at sourceware dot org> ---
That stack trace isn't really that useful.

Could you try with git gdb?
We fixed a bug like this recently.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/31465] gdb batch mode crashes when execute "gdb.post_event(lambda: gdb.execute("c"))" in a stop event handler

[ydeng339 at gatech dot edu]

https://sourceware.org/bugzilla/show_bug.cgi?id=31465

--- Comment #2 from Deng, Yunjie <ydeng339 at gatech dot edu> ---
(In reply to Tom Tromey from comment #1)
> That stack trace isn't really that useful.
> 
> Could you try with git gdb?
> We fixed a bug like this recently.

Very thanks for your timely response! I compile the gdb git project (head
acab5b12b9765d856d8618c11e4f54115396fda7) and find the crash disappeared.
However, it has a new bug --- the gdb will exit directly without continue
executing. Recall that I aim to use 'gdb.post_event(lambda: gdb.execute("c"))'
to continue the program after the stop event is finished, but it doesn't work. 

Recall that my demand is to continue executing the program after a breakpoint.
So I further try to use 'gdb.execute("c")' without 'gdb.post_event()'. That is,
the code is like this:
``` 
def on_stop(event):
    # Do some operations
    gdb.execute("c")
gdb.events.stop.connect(on_stop)
```
Although this method can continue executing the program for a while, the python
script will end with the error message "Python Exception <class
'RecursionError'>: maximum recursion depth exceeded while calling a Python
object". I use python traceback and find that the python stack is filled with
the 'on_stop()' function. It seems that a new 'on_stop()' function is executed
even the last 'on_stop()' is still in stack. Such behavior leads stack overflow
when gdb runs for a while.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/31465] gdb batch mode crashes when execute "gdb.post_event(lambda: gdb.execute("c"))" in a stop event handler

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31465

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-03-12

--- Comment #3 from Tom Tromey <tromey at sourceware dot org> ---
In your case you probably want a breakpoint with a 'stop'
method that returns False, instead.

This area can probably be improved, though.  I think what's
happening is that the initial command completes, so gdb
exits.

Maybe resurrecting -i=python is a way to go.  Unsure.

-- 
You are receiving this mail because:
You are on the CC list for the bug.