| Summary: | strict aliasing is not controllable via the option pragma | ||
|---|---|---|---|
| Product: | gcc | Reporter: | Andrew Pinski <pinskia> |
| Component: | other | Assignee: | Not yet assigned to anyone <unassigned> |
| Status: | RESOLVED FIXED | ||
| Severity: | enhancement | CC: | gcc-bugs, jakub, sandra |
| Priority: | P3 | Keywords: | documentation |
| Version: | 4.4.0 | ||
| Target Milestone: | 5.5 | ||
| Host: | Target: | ||
| Build: | Known to work: | 5.5.0, 6.1.0 | |
| Known to fail: | 5.4.0 | Last reconfirmed: | 2008-11-11 20:34:48 |
|
Description
Andrew Pinski
2008-11-11 01:23:52 UTC
Confirmed. As of current 4.9 trunk, that may have changed a bit.
but it still does not do what one would expect:
struct s
{
int x:8;
int y:24;
};
#pragma GCC push
#pragma GCC optimize ("strict-volatile-bitfields")
int foo(volatile struct s *x)
{
return x->x;
}
#pragma GCC pop
#pragma GCC push
#pragma GCC optimize ("no-strict-volatile-bitfields")
int bar(volatile struct s *x)
{
return x->x;
}
#pragma GCC pop
Both foo and bar are compiled as if -fno-strict-volatile-bitfields
was used on gcc-invocation. Whatever pragma is used last, wins...
push/pop are ignored, and does not restore the original settings.
(In reply to Bernd Edlinger from comment #2) > As of current 4.9 trunk, that may have changed a bit. > but it still does not do what one would expect: That is a different issue ... This was fixed for GCC 5.5.0 by r5-10168-g26736ac3d17699 and GCC 6.1.0+ by r6-5413-g9dc39706b4afc3 . (In reply to Andrew Pinski from comment #3) > (In reply to Bernd Edlinger from comment #2) > > As of current 4.9 trunk, that may have changed a bit. > > but it still does not do what one would expect: > > That is a different issue ... That was fixed in GCC 5.1.0 (I have not looked into what fixed it though). (In reply to Andrew Pinski from comment #4) > This was fixed for GCC 5.5.0 by r5-10168-g26736ac3d17699 and GCC 6.1.0+ by > r6-5413-g9dc39706b4afc3 . I should say that was the last patch which was needed to fix this, there was most likely others that happened beforehand. |