Menu
▾
▴
naviserver-devel
|
From: Zoran V. <zv...@ar...> - 2005-06-11 09:15:41
|
Hi
When you compile w/o debugging (i.e. with optimisation)
the GCC spits whole-lotta stuff like that:
tclthread.c:401: warning: dereferencing type-punned pointer will
break strict-aliasing rules
tclthread.c:401: warning: dereferencing type-punned pointer will
break strict-aliasing rules
The warning sems to trigger when a pointer is casted like this:
enum {
EAbsWaitIdx, EBroadcastIdx, ECreateIdx, EDestroyIdx, ESetIdx,
ESignalIdx, ETimedWaitIdx, EWaitIdx
} opt;
if (!GetArgs(interp, objc, objv, opts, 'e', ECreateIdx,
(int *) &opt, (void **) &condPtr)) {
return TCL_ERROR;
The "opt" and "condPtr" in GetArgs() are casted and it barks
at them. I cannot possibly imagine changing all code to
something like:
enum {
EAbsWaitIdx, EBroadcastIdx, ECreateIdx, EDestroyIdx, ESetIdx,
ESignalIdx, ETimedWaitIdx, EWaitIdx
} opt;
int myOpt = (int)opt;
if (!GetArgs(interp, objc, objv, opts, 'e', ECreateIdx,
&myOpt, (void **) &condPtr)) {
return TCL_ERROR;
This would mean lots of work and may introduce errors.
I've read that you can use (starting with gcc 3.3) -fno-strict-aliasing
which would remove those warnings. But also, somebody said that the
effect of the "-O2" is gone in such cases (haven't be able to verify)
What to do?
Any ideas?
Zoran
|
|
From: Vlad S. <vl...@cr...> - 2005-06-11 19:50:26
|
I went through google about this and found out that many projects like
Linux, PostgreSQL and others use -fno-strict-aliasing. And nobody could
determine if this affects optimnization, so we can use that flag as well
and in the meantime slowly change the code.
Zoran Vasiljevic wrote:
> Hi
>
> When you compile w/o debugging (i.e. with optimisation)
> the GCC spits whole-lotta stuff like that:
>
> tclthread.c:401: warning: dereferencing type-punned pointer will break
> strict-aliasing rules
> tclthread.c:401: warning: dereferencing type-punned pointer will break
> strict-aliasing rules
>
> The warning sems to trigger when a pointer is casted like this:
>
> enum {
> EAbsWaitIdx, EBroadcastIdx, ECreateIdx, EDestroyIdx, ESetIdx,
> ESignalIdx, ETimedWaitIdx, EWaitIdx
> } opt;
>
> if (!GetArgs(interp, objc, objv, opts, 'e', ECreateIdx,
> (int *) &opt, (void **) &condPtr)) {
> return TCL_ERROR;
>
> The "opt" and "condPtr" in GetArgs() are casted and it barks
> at them. I cannot possibly imagine changing all code to
> something like:
>
> enum {
> EAbsWaitIdx, EBroadcastIdx, ECreateIdx, EDestroyIdx, ESetIdx,
> ESignalIdx, ETimedWaitIdx, EWaitIdx
> } opt;
> int myOpt = (int)opt;
>
> if (!GetArgs(interp, objc, objv, opts, 'e', ECreateIdx,
> &myOpt, (void **) &condPtr)) {
> return TCL_ERROR;
>
> This would mean lots of work and may introduce errors.
>
> I've read that you can use (starting with gcc 3.3) -fno-strict-aliasing
> which would remove those warnings. But also, somebody said that the
> effect of the "-O2" is gone in such cases (haven't be able to verify)
>
> What to do?
> Any ideas?
>
> Zoran
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games. How far can you
> shotput
> a projector? How fast can you ride your desk chair down the office luge
> track?
> If you want to score the big prize, get to know the little guy. Play to
> win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> naviserver-devel mailing list
> nav...@li...
> https://lists.sourceforge.net/lists/listinfo/naviserver-devel
--
Vlad Seryakov
571 262-8608 office
vl...@cr...
http://www.crystalballinc.com/vlad/
|
|
From: Zoran V. <zv...@ar...> - 2005-06-11 20:03:48
|
Am 11.06.2005 um 21:50 schrieb Vlad Seryakov: > I went through google about this and found out that many projects > like Linux, PostgreSQL and others use -fno-strict-aliasing. And > nobody could determine if this affects optimnization, so we can use > that flag as well and in the meantime slowly change the code. > This is what I thought at the first glance but did not want to undertake anything until we clear this out. I have zero experience about what it meany to use -O switch because we always compile with symbols and w/o optimization. This is far more important (to get a decent coredump) then some (possible marginal) speed improvment. Our server itself is network/IO-bound and not CPU bound, hence we do not really care. But you might, therefore I thougt it is better to ask. Cheers Zoran |
|
From: Vlad S. <vl...@cr...> - 2005-06-11 20:12:57
|
I will need as much as possible speed for my next project where i will use NS but i do not experience with gcc optimization either. But in previous work i always wanted debugging info and sometimes -ON flags messed with gdb because of optimization. I say let's use -fno-strict-aliasing but use -fstrict-aliasing when compile for make test. That way we can clean up code. Zoran Vasiljevic wrote: > > Am 11.06.2005 um 21:50 schrieb Vlad Seryakov: > >> I went through google about this and found out that many projects >> like Linux, PostgreSQL and others use -fno-strict-aliasing. And >> nobody could determine if this affects optimnization, so we can use >> that flag as well and in the meantime slowly change the code. >> > > This is what I thought at the first glance but did not want > to undertake anything until we clear this out. > I have zero experience about what it meany to use -O switch > because we always compile with symbols and w/o optimization. > This is far more important (to get a decent coredump) then > some (possible marginal) speed improvment. Our server itself > is network/IO-bound and not CPU bound, hence we do not really > care. But you might, therefore I thougt it is better to ask. > > Cheers > Zoran > > > ------------------------------------------------------- > This SF.Net email is sponsored by: NEC IT Guy Games. How far can you > shotput > a projector? How fast can you ride your desk chair down the office luge > track? > If you want to score the big prize, get to know the little guy. Play to > win an NEC 61" plasma display: http://www.necitguy.com/?r=20 > _______________________________________________ > naviserver-devel mailing list > nav...@li... > https://lists.sourceforge.net/lists/listinfo/naviserver-devel -- Vlad Seryakov 571 262-8608 office vl...@cr... http://www.crystalballinc.com/vlad/ |
|
From: Stephen D. <sd...@gm...> - 2005-06-12 05:06:58
|
On 6/11/05, Zoran Vasiljevic <zv...@ar...> wrote:
> Hi
>=20
> When you compile w/o debugging (i.e. with optimisation)
> the GCC spits whole-lotta stuff like that:
>=20
> tclthread.c:401: warning: dereferencing type-punned pointer will
> break strict-aliasing rules
> tclthread.c:401: warning: dereferencing type-punned pointer will
> break strict-aliasing rules
>=20
> The warning sems to trigger when a pointer is casted like this:
>=20
> enum {
> EAbsWaitIdx, EBroadcastIdx, ECreateIdx, EDestroyIdx, ESetIdx,
> ESignalIdx, ETimedWaitIdx, EWaitIdx
> } opt;
>=20
> if (!GetArgs(interp, objc, objv, opts, 'e', ECreateIdx,
> (int *) &opt, (void **) &condPtr)) {
> return TCL_ERROR;
>=20
> The "opt" and "condPtr" in GetArgs() are casted and it barks
> at them. I cannot possibly imagine changing all code to
> something like:
>=20
> enum {
> EAbsWaitIdx, EBroadcastIdx, ECreateIdx, EDestroyIdx, ESetIdx,
> ESignalIdx, ETimedWaitIdx, EWaitIdx
> } opt;
> int myOpt =3D (int)opt;
>=20
> if (!GetArgs(interp, objc, objv, opts, 'e', ECreateIdx,
> &myOpt, (void **) &condPtr)) {
> return TCL_ERROR;
>=20
> This would mean lots of work and may introduce errors.
>=20
> I've read that you can use (starting with gcc 3.3) -fno-strict-aliasing
> which would remove those warnings. But also, somebody said that the
> effect of the "-O2" is gone in such cases (haven't be able to verify)
>=20
> What to do?
> Any ideas?
For much of the command parsing, I think something like this is fine:
int opt;
static CONST char *opts[] =3D {
"cleanup", "list", "create", "put", "get", NULL
};
enum {
CCleanupIdx, CListIdx, CCreateIdx, CPutIdx, CGetIdx
};
if (Tcl_GetIndexFromObj(interp, objv[1], opts, "option", 0, &opt)
!=3D TCL_OK) {
return TCL_ERROR;
}
|
|
From: Zoran V. <zv...@ar...> - 2005-06-12 09:51:17
|
Am 12.06.2005 um 07:06 schrieb Stephen Deasey:
>
>
> For much of the command parsing, I think something like this is fine:
>
> int opt;
>
> static CONST char *opts[] = {
> "cleanup", "list", "create", "put", "get", NULL
> };
> enum {
> CCleanupIdx, CListIdx, CCreateIdx, CPutIdx, CGetIdx
> };
>
> if (Tcl_GetIndexFromObj(interp, objv[1], opts, "option", 0, &opt)
> != TCL_OK) {
> return TCL_ERROR;
> }
>
>
Yes. Indeed. We should gradually go and fix those.
I will do that while walking thru. You can do as well
if at the right place.
Zoran
|
|
From: Zoran V. <zv...@ar...> - 2005-06-13 09:10:06
|
Am 12.06.2005 um 11:51 schrieb Zoran Vasiljevic: > > Yes. Indeed. We should gradually go and fix those. > I will do that while walking thru. You can do as well > if at the right place. Gradually or not gradually, I had some spare time yesterday afternoon and have fixed all of those. We can consider this one solved. Cheers Zoran |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X