5.3 problems on RH7.2

5.3 problems on RH7.2

[Evan Lavelle]

Hi - I've been using 5.3 for a couple of months, and it's generally 
working very well, but I have noticed a few problems. It would be nice 
to get some feedback on this - are these known problems, or am I doing 
something wrong?

This is on RH7.2, with gcc 3.3. I had the same issues with gcc 2.95. 
These are large-ish static executables (10-15MB), linked against 
pthreads and gtk 1.2, although the crashing code is actually 
single-threaded and doesn't use gtk. I've done a sanity check against 
ddd using the version of gdb in insight's bin directory, and it doesn't 
have these problems (although it does have lots of its own problems, of 
course).

1) I occasionally get an "Error: Can't read '_twin': no such variable" 
message. Insight crashes at this point, and has to be killed. I got a 
lot of these when debugging multi-threaded code (basically, I have to 
use ddd for multi-threaded), but I still get it even when there's only 
one thread.

2) I quite frequently get an "Error: bad text index '4'" message when 
opening a new source file. I can normally ignore this, although it does 
seem that Insight occasionally crashes some time afterwards.

3) Stack  tracing frequently doesn't work, particularly if you're 
debugging a core dump. The trace window doesn't show all the stack 
levels, and clicking on a given level takes you to the wrong level.

4) ddd also has this problem, so it's presumably gdb-related. Just how 
much can you change an executable, and expect the result to still be 
debuggable? 'run' recognises that the executable has changed, and 
reloads the symbols, but on maybe one occasion in 10 gdb will crash soon 
afterwards.

Apart from that, it's great... honest :). Certainly a lot better than ddd.

Thanks

Evan
________________________________________________________
E.M. Lavelle
Riverside Machines Ltd.  mailto:anti.spam1@dsl.pipex.com
The 'From' and 'mailto' fields contained a valid address
at the time at which this mail was posted.
________________________________________________________

Re: 5.3 problems on RH7.2

[Keith Seitz]

On Fri, 2003-08-01 at 01:39, Evan Lavelle wrote:
> This is on RH7.2, with gcc 3.3. I had the same issues with gcc 2.95. 
> These are large-ish static executables (10-15MB), linked against 
> pthreads and gtk 1.2, although the crashing code is actually 
> single-threaded and doesn't use gtk. I've done a sanity check against 
> ddd using the version of gdb in insight's bin directory, and it doesn't 
> have these problems (although it does have lots of its own problems, of 
> course).

First things first: I would recommend trying a newer version of Insight
-- especially if you're using gcc 3.3 and threads on linux. Although it
is unlikely that you will be running the newest threading model (NPTL),
if you are, the older gdb will not work at all.

> 1) I occasionally get an "Error: Can't read '_twin': no such variable" 
> message. Insight crashes at this point, and has to be killed. I got a 
> lot of these when debugging multi-threaded code (basically, I have to 
> use ddd for multi-threaded), but I still get it even when there's only 
> one thread.

I believe that this has been fixed. I don't know where or remember what
it was, though.

> 2) I quite frequently get an "Error: bad text index '4'" message when 
> opening a new source file. I can normally ignore this, although it does 
> seem that Insight occasionally crashes some time afterwards.

I think this has been mitigated. It may still occur, but I don't see it
very often. If you do see it with 6.0, I would appreciate a test case.

> 3) Stack  tracing frequently doesn't work, particularly if you're 
> debugging a core dump. The trace window doesn't show all the stack 
> levels, and clicking on a given level takes you to the wrong level.

Most likely a gdb probelm, as you indicate. When this happens, open a
console window and type "bt". Does that work? If it does, it's an
insight problem. If it doesn't, then it's a gdb problem.

Unless there is a real need to stick to 5.3, I encourage you to update
to 6.0 (unofficial), which is due for release sometime soon. You can
grab it by checking out the gdb_6_0-branch of insight.

Keith

Re: 5.3 problems on RH7.2

[Evan Lavelle]

Keith Seitz wrote:

> First things first: I would recommend trying a newer version of Insight
> -- especially if you're using gcc 3.3 and threads on linux. Although it
> is unlikely that you will be running the newest threading model (NPTL),
> if you are, the older gdb will not work at all.

Thanks - I've installed 6_0_branch. I used a test case in which Insight 
5.3 hung up when it entered a particular routine; unfortunately, 6.0 
does the same, although it does give some diagnostics. Stack trace 
below. I had to copy this out by hand - is there some way to get the 
information out of the stack trace window?

----------------------------------------------------
can't read "var_to_items(var41)": no such element in array
     while executing
"lassign $var_to_items($var) nam typ val"
     (object "::.localswin0.localswin.tree" method 
"::VarTree::update_var" body line 3)
     invoked from within
"update_var $var $enabled $check"
     (object "::.localswin0.localswin.tree" method "::VarTree::update" 
body line 16)
     invoked from within
".localswin0.localswin.tree update"
     (in namespace inscope "::LocalsWin" script line 1)
     invoked from within
"namespace inscope ::LocalsWin {.localswin0.localswin.tree update}"
     ("after" script)errorCode is NONE
------------------------------------------------------

I also had an exception when starting up Insight on this test program:

------------------------------------------------------
getting servant data 0x8200800 from window 0x816a4c8
GNOME Help-DEBUG: CORBA user exception: 
IDL:omg.org/CosNaming/NamingContext/NotFound:1.0.
------------------------------------------------------

Thanks

Evan