`imb-kat --test-type SNOW3G` fails when compiled with `SAFE_DATA=OFF` and gcc15

[michalsieron]

Environment:

• OS: Arch Linux
• CC: gcc-15.2.1 (gcc-14.3.1 works just fine)
• CPU: i5-1145G7

Steps to reproduce:

1. git checkout v1.5 # or main, issue reproduces for both
2. cmake -B build -S . -GNinja -DSAFE_DATA=OFF
3. ninja -C build/ -j8
4. ./build/test/kat-app/imb-kat --test-type SNOW3G

Expected results:

Detected library version: 1.5.0
Tool version: 1.5.0
Detected hardware features:
        SHANI        : OK
        AESNI        : OK
        PCLMULQDQ    : OK
        CMOV         : OK
        SSE4.2       : OK
        AVX          : OK
        AVX2         : OK
        AVX512(SKX)  : OK
        VAES         : OK
        VPCLMULQDQ   : OK
        GFNI         : OK
        AVX512-IFMA  : OK
        AVX-IFMA     : n/a
        BMI2         : OK
        HYBRID-CORE  : n/a
AESNI Emulation is not enabled. Skipping NOAESNI test.
SELF-TEST: PASS
[INFO] [ARCH] using SSE interface [SSE-SHANI-GFNI]
[INFO] [ALGO] SNOW3G-UEA2 PASS
[INFO] [ALGO] SNOW3G-UIA2 PASS
SELF-TEST: PASS
[INFO] [ARCH] using AVX interface [AVX]
[INFO] [ALGO] SNOW3G-UEA2 PASS
[INFO] [ALGO] SNOW3G-UIA2 PASS
SELF-TEST: PASS
[INFO] [ARCH] using AVX2 interface [AVX2]
[INFO] [ALGO] SNOW3G-UEA2 PASS
[INFO] [ALGO] SNOW3G-UIA2 PASS
SELF-TEST: PASS
[INFO] [ARCH] using AVX512 interface [AVX512-VAES-GFNI-VCLMUL]
[INFO] [ALGO] SNOW3G-UEA2 PASS
[INFO] [ALGO] SNOW3G-UIA2 PASS
Test completed: PASS

Actual results:

Detected library version: 1.5.0
Tool version: 1.5.0
Detected hardware features:
	SHANI        : OK
	AESNI        : OK
	PCLMULQDQ    : OK
	CMOV         : OK
	SSE4.2       : OK
	AVX          : OK
	AVX2         : OK
	AVX512(SKX)  : OK
	VAES         : OK
	VPCLMULQDQ   : OK
	GFNI         : OK
	AVX512-IFMA  : OK
	AVX-IFMA     : n/a
	BMI2         : OK
	HYBRID-CORE  : n/a
AESNI Emulation is not enabled. Skipping NOAESNI test.
SELF-TEST: PASS
[INFO] [ARCH] using SSE interface [SSE-SHANI-GFNI]
IMB_SNOW3G_F8_4_BUFFER(Enc) vector:2 buffer:0
Actual::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x4F 0x6D 0x7D 

Expected::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x60 0x5E 0x41 

IMB_SNOW3G_F8_8_BUFFER(Enc) vector:2 buffer:0
Actual::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0xE1 0x6D 0x42 

Expected::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x60 0x5E 0x41 

[INFO] [ALGO] SNOW3G-UEA2 FAIL
[INFO] [ALGO] SNOW3G-UIA2 PASS
SELF-TEST: PASS
[INFO] [ARCH] using AVX interface [AVX]
IMB_SNOW3G_F8_4_BUFFER(Enc) vector:2 buffer:0
Actual::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x62 0x79 0x21 

Expected::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x60 0x5E 0x41 

IMB_SNOW3G_F8_8_BUFFER(Enc) vector:2 buffer:0
Actual::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x11 0xC4 0x07 

Expected::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x60 0x5E 0x41 

[INFO] [ALGO] SNOW3G-UEA2 FAIL
[INFO] [ALGO] SNOW3G-UIA2 PASS
SELF-TEST: PASS
[INFO] [ARCH] using AVX2 interface [AVX2]
IMB_SNOW3G_F8_4_BUFFER(Enc) vector:2 buffer:0
Actual::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x62 0x79 0x21 

Expected::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x60 0x5E 0x41 

IMB_SNOW3G_F8_8_BUFFER(Enc) vector:3 buffer:0
Actual::
0xBA 0x0F 0x31 0x30 0x03 0x34 0xC5 0x6B 0xD8 0x86 0xBA 0x64 0xE2 0x33 0x4B 

Expected::
0xBA 0x0F 0x31 0x30 0x03 0x34 0xC5 0x6B 0x52 0xA7 0x49 0x7C 0xBA 0xC0 0x46 

snow3g_f8_8_multi_buffer(Enc) vector:1 buffer:0
Actual::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0x04 0x3F 0x07 0x00 0x30 0x59 0x02 0x27 
0x7F 0x31 0x97 0x1E 0xED 0x01 0xF1 0x16 0xA4 0xBF 0x28 0x15 0xF6 0xC4 0x25 0x27 


Expected::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0xE0 0xA1 0x23 0x77 0xEA 0x00 0xB3 0x7D 
0x47 0xC6 0xA0 0xBA 0x06 0x30 0x91 0x55 0x08 0x6A 0x85 0x9C 0x43 0x41 0xB3 0x7C 


IMB_SNOW3G_F8_N_BUFFER_MULTIKEY(Enc) validate_snow3g_f8_n_blocks_linear_mkeys nb_packets:7 vector: 0
Actual::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0xDB 0x64 0x8B 0x72 0xD7 0x7B 0x3D 0x98 
0xC7 0xD8 0xA1 0xD2 0x8E 0x9C 0x96 0x8D 0xE9 0xFE 0xCA 0x92 0x07 0x60 0xB0 0x12 


Expected::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0xE0 0xA1 0x23 0x77 0xEA 0x00 0xB3 0x7D 
0x47 0xC6 0xA0 0xBA 0x06 0x30 0x91 0x55 0x08 0x6A 0x85 0x9C 0x43 0x41 0xB3 0x7C 


IMB_SNOW3G_F8_N_BUFFER(Enc) , vector:0 buffer: 7
Actual::
0xE3 0x96 0x7F 0x63 0x88 0x74 0x2D 0x7C 0xF5 0xB1 0xA5 0xA6 0xB8 0x97 0xDB 0xBC 
0x2D 0xFA 0x3D 0x55 0x47 0xD4 0x05 0x1C 0x91 0x13 0x0A 0x55 0x18 0x25 0xB5 0x1C 


Expected::
0xE3 0x96 0x7F 0x63 0x88 0x74 0x2D 0x7C 0x3B 0x3D 0x5A 0x08 0xC0 0x8D 0x1A 0xC2 
0xA1 0xF2 0x38 0x96 0xC7 0x41 0x1B 0xC1 0xDE 0xAC 0x24 0x57 0x9C 0x45 0x0D 0xB0 


[INFO] [ALGO] SNOW3G-UEA2 FAIL
[INFO] [ALGO] SNOW3G-UIA2 PASS
SELF-TEST: PASS
[INFO] [ARCH] using AVX512 interface [AVX512-VAES-GFNI-VCLMUL]
IMB_SNOW3G_F8_4_BUFFER(Enc) vector:2 buffer:0
Actual::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x62 0x79 0x21 

Expected::
0x8C 0xEB 0xA6 0x29 0x43 0xDC 0xED 0x3A 0x09 0x90 0xB0 0x6E 0xA1 0xB0 0xA2 0xC4 
0xFB 0x3C 0xED 0xC7 0x1B 0x36 0x9F 0x42 0xBA 0x64 0xC1 0xEB 0x66 0x65 0xE7 0x2A 
0xA1 0xC9 0xBB 0x0D 0xEA 0xA2 0x0F 0xE8 0x60 0x58 0xB8 0xBA 0xEE 0x2C 0x2E 0x7F 
0x0B 0xEC 0xCE 0x48 0xB5 0x29 0x32 0xA5 0x3C 0x9D 0x5F 0x93 0x1A 0x3A 0x7C 0x53 
0x22 0x59 0xAF 0x43 0x25 0xE2 0xA6 0x5E 0x30 0x84 0xAD 0x5F 0x6A 0x51 0x3B 0x7B 
0xDD 0xC1 0xB6 0x5F 0x0A 0xA0 0xD9 0x7A 0x05 0x3D 0xB5 0x5A 0x88 0xC4 0xC4 0xF9 
0x60 0x5E 0x41 

IMB_SNOW3G_F8_8_BUFFER(Enc) vector:3 buffer:0
Actual::
0xBA 0x0F 0x31 0x30 0x03 0x34 0xC5 0x6B 0xD8 0x86 0xBA 0x64 0xE2 0x33 0x4B 

Expected::
0xBA 0x0F 0x31 0x30 0x03 0x34 0xC5 0x6B 0x52 0xA7 0x49 0x7C 0xBA 0xC0 0x46 

snow3g_f8_8_multi_buffer(Enc) vector:1 buffer:0
Actual::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0x04 0x3F 0x75 0xC0 0x30 0x5E 0xDA 0x77 
0x8C 0x5F 0xA7 0x1A 0xBE 0xED 0x46 0x17 0x8F 0x23 0xE0 0xA3 0xDD 0xF9 0x59 0xCC 


Expected::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0xE0 0xA1 0x23 0x77 0xEA 0x00 0xB3 0x7D 
0x47 0xC6 0xA0 0xBA 0x06 0x30 0x91 0x55 0x08 0x6A 0x85 0x9C 0x43 0x41 0xB3 0x7C 


IMB_SNOW3G_F8_N_BUFFER_MULTIKEY(Enc) validate_snow3g_f8_n_blocks_linear_mkeys nb_packets:7 vector: 0
Actual::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0xDB 0x64 0x8B 0x72 0xD7 0x7B 0x3D 0x98 
0xC7 0xD8 0xA1 0xD2 0x8E 0x9C 0x96 0x8D 0xE9 0xFE 0xCA 0x92 0x07 0x60 0xB0 0x12 


Expected::
0x5D 0x5B 0xFE 0x75 0xEB 0x04 0xF6 0x8C 0xE0 0xA1 0x23 0x77 0xEA 0x00 0xB3 0x7D 
0x47 0xC6 0xA0 0xBA 0x06 0x30 0x91 0x55 0x08 0x6A 0x85 0x9C 0x43 0x41 0xB3 0x7C 


IMB_SNOW3G_F8_N_BUFFER(Enc) , vector:0 buffer: 7
Actual::
0xE3 0x96 0x7F 0x63 0x88 0x74 0x2D 0x7C 0xF5 0xB1 0xA5 0xA6 0xB8 0x97 0xDB 0xBC 
0x2D 0xFA 0x3D 0x55 0x47 0xD4 0x05 0x1C 0x91 0x13 0x0A 0x55 0x18 0x25 0xB5 0x1C 


Expected::
0xE3 0x96 0x7F 0x63 0x88 0x74 0x2D 0x7C 0x3B 0x3D 0x5A 0x08 0xC0 0x8D 0x1A 0xC2 
0xA1 0xF2 0x38 0x96 0xC7 0x41 0x1B 0xC1 0xDE 0xAC 0x24 0x57 0x9C 0x45 0x0D 0xB0 


[INFO] [ALGO] SNOW3G-UEA2 FAIL
[INFO] [ALGO] SNOW3G-UIA2 PASS
Test completed: FAIL

[michalsieron]

The issue exists only with -O3 enabled.

I also identified following WA, which is of course not a solution, but might help with identifying the problem:

diff --git a/lib/include/snow3g_common.h b/lib/include/snow3g_common.h
index e6e960a8..9fea6834 100644
--- a/lib/include/snow3g_common.h
+++ b/lib/include/snow3g_common.h
@@ -2463,6 +2463,7 @@ SNOW3G_F8_4_BUFFER(const snow3g_key_schedule_t *pHandle, const void *pIV1, const
 {
         const size_t num_lanes = 4;
         snow3gKeyState4_t ctx;
+        CLEAR_MEM(&ctx, sizeof(ctx));
         uint32_t lenInBytes[4];
         uint8_t *pBufferOut[4];
         const uint8_t *pBufferIn[4];
@@ -2590,6 +2591,7 @@ snow3g_8_buffer_ks_32_8_multi(const snow3g_key_schedule_t *const pKey[], const v
         uint8_t *tBufferOut[8];
         uint32_t tLenInBytes[8];
         snow3gKeyState8_t ctx;
+        CLEAR_MEM(&ctx, sizeof(ctx));
         const uint32_t bytes = length_find_min(lengthInBytes, num_lanes);
         uint32_t bytes_left = bytes & (~(small_block_size - 1));
         size_t i;
@@ -2712,6 +2714,7 @@ snow3g_8_buffer_ks_32_8(const snow3g_key_schedule_t *pKey, const void *const IV[
         const size_t small_block_size = SNOW3G_8_BYTES;
         const uint32_t bytes = length_find_min(lengthInBytes, num_lanes);
         snow3gKeyState8_t ctx;
+        CLEAR_MEM(&ctx, sizeof(ctx));
         uint32_t i, bytes_left = bytes & (~(small_block_size - 1));
 
         /* Initialize the schedule from the IV */

[michalsieron]

Actually it doesn't have to be CLEAR_MEM that is called. Seems that any function that is not inlined and takes &ctx as an argument will work as this will force gcc to choose a different optimization (even printf was enough).

Most probably something to report at gcc upstream, but would need a smaller reproduction code.

[tkanteck]

Thanks for reporting the problem. We'll have a look into it.

[BKudlik]

I bisected gcc source and found the commit after which the issue becomes visible:

commit 1e0ae1f52741f7e0133661659ed2d210f939a398
Author: Richard Biener rguenther@suse.de
Date: Wed May 15 18:32:37 2024 +0200

tree-optimization/79958 - make DSE track multiple paths

DSE currently gives up when the path we analyze forks.  This leads
to multiple missed dead store elimination PRs.  The following fixes
this by recursing for each path and maintaining the visited bitmap
to avoid visiting CFG re-merges multiple times.  The overall cost
is still limited by the same bound, it's just more likely we'll hit
the limit now.  The patch doesn't try to deal with byte tracking
once a path forks but drops info on the floor and only handling
fully dead stores in that case.

        PR tree-optimization/79958
        PR tree-optimization/109087
        PR tree-optimization/100314
        PR tree-optimization/114774
        * tree-ssa-dse.cc (dse_classify_store): New forwarder.
        (dse_classify_store): Add arguments cnt and visited, recurse
        to track multiple paths when we end up with multiple defs.

        * gcc.dg/tree-ssa/ssa-dse-48.c: New testcase.
        * gcc.dg/tree-ssa/ssa-dse-49.c: Likewise.
        * gcc.dg/tree-ssa/ssa-dse-50.c: Likewise.
        * gcc.dg/tree-ssa/ssa-dse-51.c: Likewise.
        * gcc.dg/graphite/pr80906.c: Avoid DSE of last data reference
        in loop.
        * g++.dg/ipa/devirt-24.C: Adjust for extra DSE.
        * g++.dg/warn/Wuninitialized-pr107919-1.C: Use more important
        -O2 optimization level, -O1 regresses.

Compiling the library with -fno-tree-dse makes the test pass, however I'm not sure yet of performance penalties it might have.