* strict alias analysis question
@ 2003-10-14 17:02 Manfred Spraul
2003-10-14 20:58 ` Mike Stump
0 siblings, 1 reply; 2+ messages in thread
From: Manfred Spraul @ 2003-10-14 17:02 UTC (permalink / raw)
To: gcc
Hi,
the latest gcc versions generate warnings about potentially dangerous
pointer casts if strict alias analysis is enabled.
Right now two groups of warnings are generated: several from the
inlined MemSet macros:
#define MemSetLoop(start, val, len) \
do \
{ \
int32 * _start = (int32 *) (start); \
int32 * _stop = (int32 *) ((char *) \
_start + (Size) (len)); \
\
while (_start < _stop) \
*_start++ = 0; \
} while (0)
And a few from structures with common headers:
struct hdr {
...;
}
struct b {
struct hdr info;
...
};
void handler(struct hdr* parm)
{
struct b* ptr = (struct b*)parm;
}
Does MemSet break the rules for strict alias analysis? And what about
manual base classes? The actual use it a bit longer (the cast to (Node
*) in src/backend/commands/tablecmds.c generates the warning).
--
Manfred
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: strict alias analysis question
2003-10-14 17:02 strict alias analysis question Manfred Spraul
@ 2003-10-14 20:58 ` Mike Stump
0 siblings, 0 replies; 2+ messages in thread
From: Mike Stump @ 2003-10-14 20:58 UTC (permalink / raw)
To: Manfred Spraul; +Cc: gcc
On Tuesday, October 14, 2003, at 09:52 AM, Manfred Spraul wrote:
> the latest gcc versions generate warnings about potentially dangerous
> pointer casts if strict alias analysis is enabled.
> Right now two groups of warnings are generated: several from the
> inlined MemSet macros:
>
> #define MemSetLoop(start, val, len) \
> do \
> { \
> int32 * _start = (int32 *) (start); \
> int32 * _stop = (int32 *) ((char *) \
> _start + (Size) (len)); \
> \
> while (_start < _stop) \
> *_start++ = 0; \
> } while (0)
> Does MemSet break the rules for strict alias analysis?
No, it'll work fine for all int32 arrays. Will it work for anything
else, no, sorry. Well, it might work for a few other things, like char
arrays that are a multiple of around 4 bytes... I'd recommend reading
up, the standard is fairly clear and easy to read in this area.
Hint, use memcpy/memmove instead, it works.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2003-10-14 18:21 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-10-14 17:02 strict alias analysis question Manfred Spraul
2003-10-14 20:58 ` Mike Stump
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).