[Bug tree-optimization/65709] [5 Regression] Bad code for LZ4 decompression with -O3 on x86_64

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65709

--- Comment #3 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
BTW clang's output is more informative:

test.c:285:29: runtime error: load of misaligned address 0x7fe669eee001 for
type 'U64' (aka 'unsigned long'), which requires 8 byte alignment
0x7fe669eee001: note: pointer points here
<memory cannot be printed>
test.c:200:16: runtime error: load of misaligned address 0x7fe669eee009 for
type 'U16' (aka 'unsigned short'), which requires 2 byte alignment
0x7fe669eee009: note: pointer points here
 01 02 01  00 01 00 f1 04 02 00 02  00 00 00 01 00 01 0b 98  00 00 00 34 00 00
91 50  1c 00 f1 00 34
              ^ 
test.c:285:13: runtime error: store to misaligned address 0x000002020021 for
type 'U64' (aka 'unsigned long'), which requires 8 byte alignment
0x000002020021: note: pointer points here
 00 00 00  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  00 00 00 00 00 00
00 00  00 00 00 00 00
              ^ 
test.c:273:25: runtime error: load of misaligned address 0x000002020052 for
type 'U32' (aka 'unsigned int'), which requires 4 byte alignment
0x000002020052: note: pointer points here
 00 34  00 00 00 00 00 00 04 00  11 05 00 20 00 05 00 00  00 00 00 00 00 00 00
00  00 00 00 00 00 00
              ^ 
test.c:273:9: runtime error: store to misaligned address 0x000002020056 for
type 'U32' (aka 'unsigned int'), which requires 4 byte alignment
0x000002020056: note: pointer points here
 00 00 00 00 04 00  11 05 00 20 00 05 00 00  00 00 00 00 00 00 00 00  00 00 00
00 00 00 00 00  00 00
             ^ 

e.g. "type 'const void'" vs. "type 'U32' (aka 'unsigned int')"