pbemu

User-mode emulation for PocketBook 6.1-6.11 firmware on Linux desktop (podman + Wayland)

Fork 1

feat(shim): log repaint events to /var/log/repaint.log #20

Closed

dgalli.tngl.sh 3mo ago

targeting main from dgalli.tngl.sh/pbemu:feat/repaint-logging

Closed opened by and assigned to

dgalli.tngl.sh 3mo ago targeting main from dgalli.tngl.sh/pbemu:feat/repaint-logging

Agent had a lot of trouble of correctly detecting where to many repaints where made. I added this log to made debugging for the agent easier.

Every wrapped libinkview refresh call appends one line to /var/log/repaint.log (host view: /.live/var/log/repaint.log) so app authors can audit full-screen vs partial refresh frequency. Line format: FULL | RECT x y w h [flags|bw|areas], plus a per-process SUMMARY on exit (full/rect/near-full counts). Each line is a single write() on an O_APPEND fd, so concurrent guest processes never interleave. Adds the current_process_basename() helper (process_identity.c) used for the proc column; documented in shim README and AGENTS.md.

Commits 1

feat(shim): log repaint events to /var/log/repaint.log

+247 -1

Comments

v0 3mo ago

devopps.de 3mo ago

I believe shim is the wrong place for this kind of logging. It would make sense for debugging and fixing update bugs in stock firmware via binary patches, but if you are developing your own application, debug logging belongs there - this way you can also log function name and other context to find out where exactly the unwanted repaints are coming from.

dgalli.tngl.sh 3mo ago

I see your point.

The idea behind it was that this is a point where the emulator differs from a real device, since refreshes are always instant in the framebuffer.

Since this is a feature that wouldn’t need logging on a real device, as you can simply see the result on the screen, I still think it could make sense to have this inside the emulator.

devopps.de 3mo ago

As far as I know each update goes through UpdateEx, which in turn calls hw_update, which then writes to /dev/kmsg - if it's available (it's not at the moment - needs some setup):

  __stream = fopen64("/dev/kmsg","w");
  if (__stream != (FILE *)0x0) {
    _Var3 = getpid();
    fprintf(__stream,"U[%d] ",_Var3);
    _Var3 = getpid();
    fprintf(__stream,"HWUPD: pid=%d (%d,%d,%d,%d) wf=%d mode=%d wait=%d/%d\n",_Var3,param_1,param_2,
            param_3,param_4,uVar13,uVar10,DAT_00193ee4,iVar2);
    fclose(__stream);
  }

I think a better approach would be to make sure that /dev/kmsg is available in the emulator, and then these logs should appear there. You'd need to figure out correspondence of numeric wf to actual waveform names, but that's doable.

Versions