http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54971
Bug #: 54971
Summary: SRA pessimizes debug info by not creating debug stmts
for fields without replacements
Classification: Unclassified
Product: gcc
Version: 4.8.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: debug
AssignedTo: unassig...@gcc.gnu.org
ReportedBy: ja...@gcc.gnu.org
CC: jamb...@gcc.gnu.org
Consider the testcase in PR54970, or the following one.
int
main ()
{
int a[] = { 1, 2, 3 };
int *p = a + 2;
int *q = a + 1;
asm volatile ("nop");
*p += 10;
asm volatile ("nop");
*q += 10;
asm volatile ("nop");
__builtin_memcpy (&a, (int [3]) { 4, 5, 6 }, sizeof (a));
asm volatile ("nop");
*p += 10;
asm volatile ("nop");
*q += 10;
asm volatile ("nop");
return 0;
}
For a[1] and a[2] replacements are created and we eventually get corret debug
info, tracking the value changes of those array elements.
But for a[0] SRA decides not to create a replacement, as it seems to be only
assigned and never read (in PR54970 testcase just once, in the above testcase
twice).
Would it be possible for MAY_HAVE_DEBUG_STMTS to create also replacements for
those fields/elements that are only ever assigned, and at the places they used
to be assigned instead of setting the replacement to the value it should have
create a debug bind stmt? In the above testcase that would be at the place
of former a[0] = 1; add DEBUG a$0 => 1 stmt (where a$0's DECL_DEBUG_EXPR would
be a[0]), and at the point of the MEM[&a, 0] = MEM[something, 0] assignment
DEBUG a$0 => 4 ?
