Mathieu Tarral

From https://github.com/radare/radare2/blob/b25653ec2ae48bad9a554753d0c21ce86b0e7e0a/libr/debug/p/debug_native.c#L1523: ~~~C /* * set or unset breakpoints... * * we only handle the case for hardware breakpoints here. otherwise, * we let the caller handle the work. */...

I found a documentation regarding debugger internals: https://github.com/radare/radare2/blob/master/doc/debug-internals.md

Yes the caller is handling writing the software breakpoint: https://github.com/radare/radare2/blob/master/libr/bp/bp_io.c#L49

@radare interesting, the output is always almost a magic size: either this: ~~~ INFO:root:output len: 4095 ~~~ or this value: ~~~ INFO:root:output len: 8191 ~~~ =)

@radare no, i'm not printing any buffer, just strings. this works fine with `r2`.

For example, if i try to intercept the kernel process (the one which is the most frequently scheduled), the script is working and i have `pd 10` output: `len(output)` here...

@quim0 i'm developing a debugger plugin based on virtual machine introspection: Here: https://github.com/Wenzel/radare2-extras/blob/python/vmi/r2vmi.py#L47 Ask me if you want to test a script or a behavior :)

hmm, if you would print a nullbyte, the buffer length would not be a magic number like `4095` in my opinion. Thank you for investigating. I will take a look...

Hello, you know it's the week-end don't you :) but anyway: OS: Ubuntu 16.04 stable Python 3.5.2 I made a video to demonstrate how it happens: https://drive.google.com/file/d/1D8Zj2MB35y3lq7MxcpVgC4CJmvDSxkdH/view?usp=sharing Also note that...

@radare i made some progress on this bug, it's still here, and i have a Dockerfile for you to reproduce: ~~~Dockerfile FROM radare/radare2 USER root RUN apt-get update && apt-get...