SIGTRAP problem (and other things)

SIGTRAP problem (and other things)

[Gene Smith]

I was using an older build for a while (since Mar 2008) that has been 
OK, but not perfect. However, recently I started getting, after adding 
some code to my embedded ARM application, a strange error out of the 
blue while running under insight.

It spontaneously pops up a box that looks like I have hit the "Stop" 
icon or maybe a breakpoint. This is while the application is running 
along fine. Specifically it says "! Program received signal SIGTRAP, 
Trace/breakpoint trap". This happens after a random amount of time, 
usually less than 1/2 hour with no breakpoints set.

I have tried just using command line gdb or even gdbtui with the same 
application elf and I don't have a problem -- they allow the app to run 
forever with no SIGTRAP occurring.

When you look at the PC register after the SIGTRAP, it is the address of 
the Green highlighted text on the source where you last were, but no 
breakpoints are set when this error occurs.

Other problems I have had with insight are

1. If more than 1 breakpoint is set, code starts stepping into function 
even when I am doing N. This does not occur with straight gdb/gdbtui 
(can have more than 1 bp and it N works as expected).
2. Balloons set on crashes insight after a while (this may or may not be 
fixed in new version. (n/a with straight gdb or gdbtui).
3. If you examine a value in in the console window, x 0xabcd1234, it 
show the value OK. But if you recall the function later (up/dn arrow) 
you only see the x. It looses the 0xabcd1234 and it must be entered again.
4. Pops up gdb warning in dialog boxes that must be acknowledged. One 
was quite long and thought maybe it was causing a buffer overrun and 
causing other problem. Got rid of it (something about a macro redefined 
in my code) but still have the SIGTRAP problem).

Tried a newer snapshot (only 20081021 with a manual patch would build) 
and it didn't seem to help (except possibly with the balloon crash, 
didn't see a crash with balloons on).

When insight works it is great. But guess I need to learn better command 
line gdb skills since posts by the maintainer on this list indicate that 
support for it is going away.

-gene

Re: SIGTRAP problem (and other things)

[Gene Smith]

Gene Smith wrote, On 10/30/2008 05:46 PM:
> I was using an older build for a while (since Mar 2008) that has been 
> OK, but not perfect. However, recently I started getting, after adding 
> some code to my embedded ARM application, a strange error out of the 
> blue while running under insight.
> 
> It spontaneously pops up a box that looks like I have hit the "Stop" 
> icon or maybe a breakpoint. This is while the application is running 
> along fine. Specifically it says "! Program received signal SIGTRAP, 
> Trace/breakpoint trap". This happens after a random amount of time, 
> usually less than 1/2 hour with no breakpoints set.
> 
> I have tried just using command line gdb or even gdbtui with the same 
> application elf and I don't have a problem -- they allow the app to run 
> forever with no SIGTRAP occurring.
> 
> When you look at the PC register after the SIGTRAP, it is the address of 
> the Green highlighted text on the source where you last were, but no 
> breakpoints are set when this error occurs.

I was wrong, the SIGTRAP is *not* caused by Insight. I also saw it with 
just command line GDB. However, don't see it when JTAG debugger 
disconnected. So not sure if coming from GDB or debugger.

> 
> Other problems I have had with insight are
> 
> 1. If more than 1 breakpoint is set, code starts stepping into function 
> even when I am doing N. This does not occur with straight gdb/gdbtui 
> (can have more than 1 bp and it N works as expected).
> 2. Balloons set on crashes insight after a while (this may or may not be 
> fixed in new version. (n/a with straight gdb or gdbtui).
> 3. If you examine a value in in the console window, x 0xabcd1234, it 
> show the value OK. But if you recall the function later (up/dn arrow) 
> you only see the x. It looses the 0xabcd1234 and it must be entered again.
> 4. Pops up gdb warning in dialog boxes that must be acknowledged. One 
> was quite long and thought maybe it was causing a buffer overrun and 
> causing other problem. Got rid of it (something about a macro redefined 
> in my code) but still have the SIGTRAP problem).
> 
> Tried a newer snapshot (only 20081021 with a manual patch would build) 
> and it didn't seem to help (except possibly with the balloon crash, 
> didn't see a crash with balloons on).
> 
> When insight works it is great. But guess I need to learn better command 
> line gdb skills since posts by the maintainer on this list indicate that 
> support for it is going away.
> 
> -gene
> 
> 
> 
> 
> 
>