Typing any character in the terminal causes jump to source

[Charlie-83]

This makes things pretty unusable.

The issue is cause by this change:
09c92da#diff-f73960b7024bf1aed12c42b6d3c29f67b52a1dfee2fb99cc07d937d6d340bb3dL50

The pattern $ matches every string so this code is triggered every time the input buffer is changed. I have fixed the issue by reverting this change to the actual regex for the (gdb) prompt.

I am happy to submit a PR but I don't understand why it was changed to $ in the first place. Is there something I am missing?

[sakhnik]

I remember that looking for just the prompt was fragile. Therefore, I added a custom marker ^z^z^z at some point. But it wasn't reliable in some scenarios, either.
I can't reproduce the issue, I can type any commands. Could you please give me more details? Could you investigate what's happening?

[Charlie-83]

https://asciinema.org/a/TGYF2fu0k5LYfW2BBKzjx3vy1

I need to look into this issue more to find a reproducible case but here is a video showing this issue.
I think that the issue can also be much worse depending on the config. I was getting it so that it also jumped my cursor to the window with the code which made it essentially impossible to type anything in the console; you need to change back to the terminal window after each character. I believe this had something to do with sticky_dbg_buf being false but I was changing other stuff also at the time

Edit: Confirmed that this issue is much worse when sticky_dbg_buf is false

[sakhnik]

How does the branch develop work for you?

[Charlie-83]

Testing on 0da3d19

This has fixed the issue where typing any character causes it to jump.

This has also made commands like p 123 not jump (which was the aim of my PR).

There is a problem I have noticed.

• Set a breakpoint
• Run the program. It will hit the breakpoint and jump to it
• Run the program again (run). It will hit the breakpoint and won't jump
• Run it again and it will jump
• Each time you run the program and it hits that same breakpoint, it alternates between jumping and not jumping
  https://asciinema.org/a/rZiJ4nJ3ty4cd77df3qNZWMj7

The same thing can be achieved by placing the breakpoint in a loop and using continue rather than run.

I think that the ideal behaviour is that p 123 doesn't cause a jump while continue/run will cause a jump even if it is to the same place as last time. This problem was something I fixed by adding a new reset-on-read flag to gdb_commands.py to record whether the program had actually ran or not since the last time it stopped.

[sakhnik]

When your fix is slightly adjusted, it fixes #213 too.
I'll still have to investigate how to do that same for lldb.