On September 18th, 2013, 2:40 p.m. CEST, Jan Synacek wrote:

src/account/LMI_AccountProvider.c (Diff revision 1)
59	if (init_lock_pool() == 0) {	59	const int ret = init_lock_pool();

This change is totally useless and introduces a variable that is thrown away after a few lines anyway.

On September 18th, 2013, 2:44 p.m. CEST, Robin Hack wrote:

Ok. GCC without optimalizations convinst me.
With variable:
   0x000000000040047f <+0>:	push   %rbp
   0x0000000000400480 <+1>:	mov    %rsp,%rbp
   0x0000000000400483 <+4>:	sub    $0x10,%rsp
   0x0000000000400487 <+8>:	mov    $0x0,%eax
   0x000000000040048c <+13>:	callq  0x400474 <c>
   0x0000000000400491 <+18>:	mov    %eax,-0x4(%rbp)
   0x0000000000400494 <+21>:	cmpl   $0x1,-0x4(%rbp)
   0x0000000000400498 <+25>:	jne    0x4004a1 <main+34>
   0x000000000040049a <+27>:	mov    $0x0,%eax
   0x000000000040049f <+32>:	jmp    0x4004a6 <main+39>
   0x00000000004004a1 <+34>:	mov    $0x1,%eax
   0x00000000004004a6 <+39>:	leaveq 
   0x00000000004004a7 <+40>:	retq   

Without variable:
   0x000000000040047f <+0>:	push   %rbp
   0x0000000000400480 <+1>:	mov    %rsp,%rbp
   0x0000000000400483 <+4>:	mov    $0x0,%eax
   0x0000000000400488 <+9>:	callq  0x400474 <c>
   0x000000000040048d <+14>:	cmp    $0x1,%eax
   0x0000000000400490 <+17>:	jne    0x400499 <main+26>
   0x0000000000400492 <+19>:	mov    $0x0,%eax
   0x0000000000400497 <+24>:	jmp    0x40049e <main+31>
   0x0000000000400499 <+26>:	mov    $0x1,%eax
   0x000000000040049e <+31>:	leaveq 
   0x000000000040049f <+32>:	retq   

Yeah, definitely a worth optimization at the right place...

- Jan

On September 18th, 2013, 2:28 p.m. CEST, Robin Hack wrote:

Account: Better failure check of lock pool initialization.