Malformed LLVM module about llvm-mctoll HOT 4 CLOSED

Fixed the syntax error in test provided and added to tool tests. Passes.

from llvm-mctoll.

Thanks for using the tool and for the bug report.

Can you add the gcc version, as well as the Linux distro used (may be the later is not entirely relevant, but please do so, if you can) to the details of this issue?

Using gcc 7.4.0 on Ubunbtu 18.04.04, following is the code generated for foo():

0000000000000610 <foo>:
 610:	48 89 f8             	mov    rax,rdi
 613:	eb 06                	jmp    61b <foo+0xb>
 615:	0f 1f 00             	nop    DWORD PTR [rax]
 618:	48 89 d0             	mov    rax,rdx
 61b:	48 8b 10             	mov    rdx,QWORD PTR [rax]
 61e:	48 85 d2             	test   rdx,rdx
 621:	75 f5                	jne    618 <foo+0x8>
 623:	f3 c3                	repz ret 
 625:	66 2e 0f 1f 84 00 00 	nop    WORD PTR cs:[rax+rax*1+0x0]
 62c:	00 00 00 
 62f:	90                   	nop

Unfortunately, the code generated with gcc 7.4.0 does not demonstrate the code pattern that triggers the malformed tool output reported in this issue. So, it would be very beneficial to get the gcc version that generates the code pattern triggering the failure, thus allowing for support to be added to correctly handle such a code pattern.

While the solution you propose seems reasonable, I would also like to look at the possibility of addressing it during BasicBlock creation.

Thanks!

from llvm-mctoll.

Hello,
Thanks you for the response. My machine is Fedora 31

gcc --version
gcc (GCC) 9.2.1 20190827 (Red Hat 9.2.1-1)

from llvm-mctoll.

Thanks! Since I do not have immediate access to that version of gcc, can you please provide the assembly listing of foo.c generated on your machine using the above gcc version (gcc -O3 -S foo.c should generate foo.s)? This would help with creation of a test case as well as allow for implementing a fix.

Any help is highly appreciated.

from llvm-mctoll.