diff options
Diffstat (limited to 'results/classifier/108/device/2905')
| -rw-r--r-- | results/classifier/108/device/2905 | 39 |
1 files changed, 39 insertions, 0 deletions
diff --git a/results/classifier/108/device/2905 b/results/classifier/108/device/2905 new file mode 100644 index 000000000..9a998b67c --- /dev/null +++ b/results/classifier/108/device/2905 @@ -0,0 +1,39 @@ +device: 0.941 +graphic: 0.933 +debug: 0.914 +socket: 0.898 +performance: 0.862 +vnc: 0.820 +network: 0.735 +semantic: 0.725 +files: 0.720 +permissions: 0.619 +PID: 0.581 +other: 0.494 +boot: 0.379 +KVM: 0.082 + +Windows Curses Display Infinite Loop +Description of problem: +The out-of-the-box `qemu-system-x86_64 -display curses` on Windows loops forever while displaying "VGA Blank Mode" instead of booting like `qemu-system-x86_64` does. + +This is caused by an infinite loop in the below simplified code in `curses_refresh` in `ui/curses.c`: +``` + int chr; + // ...trimmed + while (1) { + /* while there are any pending key strokes to process */ + chr = console_getch(&maybe_keycode); + + if (chr == -1) + break; + // ...trimmed + } +``` +`console_getch` has return type `wint_t`. However, on Windows, `wint_t` is `unsigned short`. Therefore when `console_getch` returns -1, the -1 value of `unsigned short` will be silently converted into the `int` value 65535. This causes `65535 == -1` to always be false, and the loop will never break. I can send a patch to qemu-devel which retypes `chr` to `wint_t` and replaces occurences of -1 with `WEOF` (an alias for `(wint_t) -1`). +Steps to reproduce: +1. Install `qemu-w64-setup-20250326.exe` Windows qemu from https://qemu.weilnetz.de/w64/2025/ +2. Run `./qemu-system-x86_64 -display curses` +3. "VGA Blank Mode" will appear on the screen forever +Additional information: + |