* unitialised error
@ 2004-07-01 12:05 Ken Foskey
2004-07-01 16:22 ` llewelly
2004-07-01 20:48 ` Falk Hueffner
0 siblings, 2 replies; 6+ messages in thread
From: Ken Foskey @ 2004-07-01 12:05 UTC (permalink / raw)
To: gcc help
The following code gives me an uninitialised warning on nCol. There is
no possible default clause for this one.
switch( nXIndex & 3 )
{
case 0 :
nCol = *pTmp >> 6;
break;
case 1 :
nCol = ( *pTmp >> 4 ) & 0x03 ;
break;
case 2 :
nCol = ( *pTmp >> 2 ) & 0x03;
break;
case 3 :
nCol = ( *pTmp++ ) & 0x03;
break;
}
--
Thanks
KenF
OpenOffice.org developer
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: unitialised error
2004-07-01 12:05 unitialised error Ken Foskey
@ 2004-07-01 16:22 ` llewelly
2004-07-01 20:48 ` Falk Hueffner
1 sibling, 0 replies; 6+ messages in thread
From: llewelly @ 2004-07-01 16:22 UTC (permalink / raw)
To: Ken Foskey; +Cc: gcc help
Ken Foskey <foskey@optushome.com.au> writes:
> The following code gives me an uninitialised warning on nCol. There is
> no possible default clause for this one.
>
> switch( nXIndex & 3 )
> {
> case 0 :
> nCol = *pTmp >> 6;
> break;
> case 1 :
> nCol = ( *pTmp >> 4 ) & 0x03 ;
> break;
>
> case 2 :
> nCol = ( *pTmp >> 2 ) & 0x03;
> break;
> case 3 :
> nCol = ( *pTmp++ ) & 0x03;
> break;
> }
What version of gcc are you using? I don't get any such warning for this
code with gcc 2.95.3, 3.3.3 or 3.4 .
I completed it like this:
int main()
{
int nXIndex;
int nCol;
int Tmp;
int *pTmp;
switch( nXIndex & 3 )
{
case 0 :
nCol = *pTmp >> 6;
break;
case 1 :
nCol = ( *pTmp >> 4 ) & 0x03 ;
break;
case 2 :
nCol = ( *pTmp >> 2 ) & 0x03;
break;
case 3 :
nCol = ( *pTmp++ ) & 0x03;
break;
}
Tmp= nCol;
}
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: unitialised error
2004-07-01 12:05 unitialised error Ken Foskey
2004-07-01 16:22 ` llewelly
@ 2004-07-01 20:48 ` Falk Hueffner
2004-07-02 5:35 ` llewelly
1 sibling, 1 reply; 6+ messages in thread
From: Falk Hueffner @ 2004-07-01 20:48 UTC (permalink / raw)
To: Ken Foskey; +Cc: gcc help
Ken Foskey <foskey@optushome.com.au> writes:
> The following code gives me an uninitialised warning on nCol. There is
> no possible default clause for this one.
>
> switch( nXIndex & 3 )
> {
> case 0 :
> nCol = *pTmp >> 6;
> break;
> case 1 :
> nCol = ( *pTmp >> 4 ) & 0x03 ;
> break;
>
> case 2 :
> nCol = ( *pTmp >> 2 ) & 0x03;
> break;
> case 3 :
> nCol = ( *pTmp++ ) & 0x03;
> break;
> }
There will always be cases where gcc guesses wrong on this; it is not
only very hard, but provably impossible to get it always right.
In this case however, it indicates an optimizer deficiency. You could
try to file a bug report about it.
By the way, please always post complete test cases and include version
and command line options.
--
Falk
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: unitialised error
2004-07-01 20:48 ` Falk Hueffner
@ 2004-07-02 5:35 ` llewelly
2004-07-02 8:03 ` Ken Foskey
0 siblings, 1 reply; 6+ messages in thread
From: llewelly @ 2004-07-02 5:35 UTC (permalink / raw)
To: Falk Hueffner; +Cc: Ken Foskey, gcc help
Falk Hueffner <hueffner@informatik.uni-tuebingen.de> writes:
> Ken Foskey <foskey@optushome.com.au> writes:
>
> > The following code gives me an uninitialised warning on nCol. There is
> > no possible default clause for this one.
> >
> > switch( nXIndex & 3 )
> > {
> > case 0 :
> > nCol = *pTmp >> 6;
> > break;
> > case 1 :
> > nCol = ( *pTmp >> 4 ) & 0x03 ;
> > break;
> >
> > case 2 :
> > nCol = ( *pTmp >> 2 ) & 0x03;
> > break;
> > case 3 :
> > nCol = ( *pTmp++ ) & 0x03;
> > break;
> > }
>
> There will always be cases where gcc guesses wrong on this; it is not
> only very hard, but provably impossible to get it always right.
However I get no such warnings for the same code, not with gcc 3.4,
3.3.3, or very old 2.95.3 .
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: unitialised error
2004-07-02 5:35 ` llewelly
@ 2004-07-02 8:03 ` Ken Foskey
2004-07-02 13:45 ` llewelly
0 siblings, 1 reply; 6+ messages in thread
From: Ken Foskey @ 2004-07-02 8:03 UTC (permalink / raw)
To: gcc help
On Fri, 2004-07-02 at 15:35, llewelly@xmission.com wrote:
> However I get no such warnings for the same code, not with gcc 3.4,
> 3.3.3, or very old 2.95.3 .
You need to turn on optimisation and -Wuninitialized
Very good practice, I have found lots of bugs this way.
--
Thanks
KenF
OpenOffice.org developer
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: unitialised error
2004-07-02 8:03 ` Ken Foskey
@ 2004-07-02 13:45 ` llewelly
0 siblings, 0 replies; 6+ messages in thread
From: llewelly @ 2004-07-02 13:45 UTC (permalink / raw)
To: Ken Foskey; +Cc: gcc help
Ken Foskey <foskey@optushome.com.au> writes:
> On Fri, 2004-07-02 at 15:35, llewelly@xmission.com wrote:
>
> > However I get no such warnings for the same code, not with gcc 3.4,
> > 3.3.3, or very old 2.95.3 .
>
> You need to turn on optimisation and -Wuninitialized
I had optimization and -Wuninitialized. Here was the full compile
line:
cd /home/llewelly/llewelly/cc_moderated/
g++-3.4 -O2 -W -Wall -Wuninitialized foskey.cc
Compilation finished at Fri Jul 2 07:41:03
(I just re-verified)
Maybe the surrounding code you didn't paste affected the diagnostic. I
completed your original example like this:
int main()
{
int nXIndex;
int nCol;
int Tmp;
int *pTmp;
switch( nXIndex & 3 )
{
case 0 :
nCol = *pTmp >> 6;
break;
case 1 :
nCol = ( *pTmp >> 4 ) & 0x03 ;
break;
case 2 :
nCol = ( *pTmp >> 2 ) & 0x03;
break;
case 3 :
nCol = ( *pTmp++ ) & 0x03;
break;
}
Tmp= nCol;
}
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2004-07-02 13:45 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-07-01 12:05 unitialised error Ken Foskey
2004-07-01 16:22 ` llewelly
2004-07-01 20:48 ` Falk Hueffner
2004-07-02 5:35 ` llewelly
2004-07-02 8:03 ` Ken Foskey
2004-07-02 13:45 ` llewelly
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).