insight/226: core dump when cursor move over class member function without debugger info.

insight/226: core dump when cursor move over class member function without debugger info.

[ching]

>Number:         226
>Category:       insight
>Synopsis:       core dump when cursor move over class member function without debugger info.
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed May 07 22:53:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     ching@coware.com
>Release:        unknown-1.0
>Organization:
>Environment:
Linux 7.3 and Solaris. 
>Description:
insight 5.2 crashed when the cursor moves over a class
member function that did not compile with debugger
information.

Platform: Redhat 7.3

Linux scooby 2.4.18-3 #1 Thu Apr 18 07:37:53 EDT 2002 i686 unknown



Here is of my observation. It looks like the problem is caused
by tk_command_loop() cannot not re-entry. So what happens is
when the error occurs in

gdb/cp-valprint.c
    123             }
    124           error ("invalid pointer to member function");
    125         }

which caused a siglongjump to gdb/top.c

    421   caught = SIGSETJMP (catch);
    422   if (!caught)
    423     val = (*func) (func_uiout, func_args);


after that it will call tk_command_loop() in gdb/gdbtk/generic/gdbtk-hooks.c
to re-enter the Tk_MainLoop. Since it is the same gdbtk_interp tcl interpeter
and it keeps evaluation the same tcl_command and finially the stack is screwed up.

>How-To-Repeat:
To reproduce:
1. unzip the carsh.bz2 file.
2. cd crash
3. invoke insight
4. file t
5. b t.cpp:60
6. run
7. move cusor over line 60 of "data_valid.read()"

after that you will see dozen of the following messages.

invalid pointer to member function

8. enter c to continue debugging
9. Segmentation fault (core dumped)
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted: