Re: [hatari-devel] ASAN issues with tests

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

This issue is real.

"screen-end" VDI test segfaults on CI MAC builds, which do not use ASAN:
------------------------------------------------
/private/var/folders/tn/f_9sf1xx5t14qm_6f83q3b840000gn/T/cirrus-ci-build/tests/mem_end/run_test.sh: 
line 30: 10800 Segmentation fault: 11  HOME="$testdir" $hatari 
--log-level fatal --sound off --bios-intercept on --fast-forward on 
--run-vbls 500 --frameskips 0 --tos none --screenshot-dir "$testdir" 
"$@" "$prg" > "$testdir/out.txt" 2>&1
Running hatari FAILED. Status=139. Hatari output:
Screen dump saved to: 
/var/folders/tn/f_9sf1xx5t14qm_6f83q3b840000gn/T/tmp.QwrOEVMg/grab0001.png
<end of output>
------------------------------------------------

See: https://cirrus-ci.com/task/5653629708271616


	- Eero

On 29.7.2022 17.33, Nicolas Pomarède wrote:
> Le 29/07/2022 à 16:20, Nicolas Pomarède a écrit :
> And for the -vdi tests, I see this :
> ==29173==ERROR: AddressSanitizer: heap-buffer-overflow on address 
> 0x7f5c0aa73840 at pc 0x0000006e5611 bp 0x7fffd2b6b690 sp 0x7fffd2b6b688
> READ of size 4 at 0x7f5c0aa73840 thread T0
>      #0 0x6e5610 in Screen_BitplaneToChunky32 
> /home/npomarede/src/hatari.git/src/screenConvert.c:205
>      #1 0x6e93c9 in ScreenConv_BitplaneLineTo32bpp 
> /home/npomarede/src/hatari.git/src/screenConvert.c:342
>      #2 0x6e93c9 in ScreenConv_BitplaneTo32bppNoZoom 
> /home/npomarede/src/hatari.git/src/screenConvert.c:453
>      #3 0x6e93c9 in Screen_ConvertWithoutZoom 
> /home/npomarede/src/hatari.git/src/screenConvert.c:680
>      #4 0x6e93c9 in Screen_GenConvert 
> /home/npomarede/src/hatari.git/src/screenConvert.c:1146
>      #5 0x6ea727 in Screen_GenDraw 
> /home/npomarede/src/hatari.git/src/screenConvert.c:1166
>      #6 0x7142bd in Video_DrawScreen 
> /home/npomarede/src/hatari.git/src/video.c:4399
>      #7 0x7142bd in Video_InterruptHandler_VBL 
> /home/npomarede/src/hatari.git/src/video.c:4577
....
> 0x7f5c0aa73840 is located 0 bytes to the right of 4251712-byte region 
> [0x7f5c0a665800,0x7f5c0aa73840)
> allocated by thread T0 here:
>      #0 0x7f5c0d6b8757 in calloc (/lib64/libasan.so.8+0xb8757)
>      #1 0x8575e0 in memory_init 
> /home/npomarede/src/hatari.git/src/cpu/memory.c:1676
>      #2 0x6fa846 in TOS_InitImage 
> /home/npomarede/src/hatari.git/src/tos.c:1135
>      #3 0x6c33b1 in Reset_ST /home/npomarede/src/hatari.git/src/reset.c:61
>      #4 0x5c6528 in Main_Init /home/npomarede/src/hatari.git/src/main.c:757
>      #5 0x5c6528 in main /home/npomarede/src/hatari.git/src/main.c:951
....
> I'm not familiar with Screen_BitplaneToChunky32(), can't tell if it's a 
> false positive due to the various casts used to convert screen.