Tudor Brindus
Tudor Brindus
Currently, tracing them fails because we can't open `/proc/$pid/exe`, but we can figure out if we're looking at a kernel thread (its parent's PID should be 2), and use vmlinux...
We can place the trigger breakpoint ~wherever we want. Currently we allow placing it only at function starts, but it should be possible to use the debug info to resolve...
We can mark them with an argument, as well as a flow marker (arrow from one frame to the other). We may need to special-case this for OCaml-only, depending on...
This would allow doing e.g. ``` $ magic-trace attach -p $(pidof foo) -trigger bar sleep 5 ``` like `perf`.
This allows capturing traces with `-trigger` symbols that get called with very high (possibly bursty) frequency, where we otherwise migt trigger before the snapshot buffer has a chance to fill...
Tested with `MAGIC_TRACE_SHARE_COMMAND=/usr/bin/echo`. Also fixed a bug where we were unconditionally serving the UI if the Perfetto base dir was specified. We lose port configurability, but it's not a big...
Sometimes it is useful to know where a function was called from. We currently don't display this information.
We cannot support arbitrary `_setjmp`/`longjmp`, but if we see a `longjmp` we can probably avoid totally breaking the trace by resetting the stack, and ignoring future `ret`s underflowing our call...
Hi, I'm running into some initialization issues while testing an app under a certain device, with the app exiting violently. I implement a backup software renderer in case OpenGL.Net is...
Hi, I have a usage of libseccomp that involves a 64-bit process setting up a filter for and then launching a 32-bit process, with `SCMP_ACT_TRACE` and using `ptrace` for syscall...