https://bugs.kde.org/show_bug.cgi?id=523257

Marcus Seyfarth <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|BACKTRACE                   |---
             Status|NEEDSINFO                   |REPORTED

--- Comment #3 from Marcus Seyfarth <[email protected]> ---
I've attached the requested backtraces. I want to point out that KWin does
**not** crash here, so there is no `kwin_wayland` coredump and DrKonqi never
pops up. The process that receives SIGSEGV is **Xwayland** — a *client* of
KWin. KWin's only fault is that it transiently sends an invalid `0×0` output
size (`wl_output.mode` / `xdg_output.logical_size`); Xwayland then feeds that
`0×0` into libxcvt_gen_mode_info()`, which (correctly, since latest libxcvt
0.1.3-git) returns `NULL`, and Xwayland dereferences it and aborts the whole X
server.

So the backtrace KDE needs is **Xwayland's** backtrace attached here (produced
in a controlled harness against the real libxcvt + the real
`hw/xwayland/xwayland-cvt.c`; it matches your journal byte-for-byte, incl. the
`mov 0x10(%rax)` / “Segmentation fault at address 0x10” form).

The call chain is:
`apply_output_change()` → `output_get_rr_modes()` → `xwayland_cvt()` →
`xwayland_modeinfo_from_cvt()` → **NULL deref of the libxcvt result**
(`xwayland-cvt.c:42/44`).

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to