/* S390 channel-to-channnel devices have type 256 */
- if (type != 1&& !strncmp(ent->d_name, "ctc", 3)&& type !=
256)
Here all the expressions are and'ed.
Test 1:
type = 1
ent->d_name = "foo"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) FAIL
type != 256 PASS
got FAIL && FAIL && PASS = FAIL
Test 2:
type = 1
end->d_name = "ctc"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) PASS
type != 256 PASS
got FAIL && PASS && PASS = FAIL
Test 3:
type = 2
end->d_name = "ctc"
type != 1 PASS
!strncmp(ent->d_name, "ctc", 3) PASS
type != 256 PASS
got PASS && PASS && PASS = PASS
+ if (type != 1 || (!strncmp(ent->d_name, "ctc", 3)&& type !=
256))
Here the last two expressions are and'ed and that result is or'ed with the
first expression.
Test 1:
type = 1
ent->d_name = "foo"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) FAIL
&& FAIL
type != 256 PASS /
got FAIL || FAIL = FAIL
Test 2:
type = 1
end->d_name = "ctc"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) PASS
&& PASS
type != 256 PASS /
got FAIL || PASS = PASS
Test 3:
type = 2
end->d_name = "ctc"
type != 1 PASS
!strncmp(ent->d_name, "ctc", 3) PASS
&& PASS
type != 256 PASS /
got PASS || PASS = PASS
continue;
new = calloc(1, sizeof(struct device));
I feel confused, but shouldn't the condition stand:
if ( type != 1 && !(!strncmp(ent->d_name, "ctc", 3) && type == 256) )
That is: ignore the same as before except for device that has type 256
and starts with "cnc"?
(original condition was: if (type != 1))
We went from FAIL, FAIL, PASS to FAIL, PASS, PASS with the addition of the
parens. Without the patch, the code requires ctc devices to be type != 1 and
type != 256, when we really need to just check that type != 256 if we have a
ctc device.
--
David Cantrell <dcantrell@redhat.com>
Red Hat / Honolulu, HI
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
/* S390 channel-to-channnel devices have type 256 */
- if (type != 1&& !strncmp(ent->d_name, "ctc", 3)&& type !=
256)
Here all the expressions are and'ed.
Test 1:
type = 1
ent->d_name = "foo"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) FAIL
type != 256 PASS
got FAIL && FAIL && PASS = FAIL
Test 2:
type = 1
end->d_name = "ctc"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) PASS
type != 256 PASS
got FAIL && PASS && PASS = FAIL
Test 3:
type = 2
end->d_name = "ctc"
type != 1 PASS
!strncmp(ent->d_name, "ctc", 3) PASS
type != 256 PASS
got PASS && PASS && PASS = PASS
+ if (type != 1 || (!strncmp(ent->d_name, "ctc", 3)&& type !=
256))
Here the last two expressions are and'ed and that result is or'ed with the
first expression.
Test 1:
type = 1
ent->d_name = "foo"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) FAIL
&& FAIL
type != 256 PASS /
got FAIL || FAIL = FAIL
Test 2:
type = 1
end->d_name = "ctc"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) PASS
&& PASS
type != 256 PASS /
got FAIL || PASS = PASS
Test 3:
type = 2
end->d_name = "ctc"
type != 1 PASS
!strncmp(ent->d_name, "ctc", 3) PASS
&& PASS
type != 256 PASS /
got PASS || PASS = PASS
continue;
new = calloc(1, sizeof(struct device));
I feel confused, but shouldn't the condition stand:
if ( type != 1 && !(!strncmp(ent->d_name, "ctc", 3) && type == 256) )
That is: ignore the same as before except for device that has type 256
and starts with "cnc"?
(original condition was: if (type != 1))
We went from FAIL, FAIL, PASS to FAIL, PASS, PASS with the addition of the
parens. Without the patch, the code requires ctc devices to be type != 1 and
type != 256, when we really need to just check that type != 256 if we have a
ctc device.
Oh, and I meant ack by all that.
--
David Cantrell <dcantrell@redhat.com>
Red Hat / Honolulu, HI
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
/* S390 channel-to-channnel devices have type 256 */
- if (type != 1&& !strncmp(ent->d_name, "ctc", 3)&&
type != 256)
Here all the expressions are and'ed.
Test 1:
type = 1
ent->d_name = "foo"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) FAIL
type != 256 PASS
got FAIL && FAIL && PASS = FAIL
Test 2:
type = 1
end->d_name = "ctc"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) PASS
type != 256 PASS
got FAIL && PASS && PASS = FAIL
Test 3:
type = 2
end->d_name = "ctc"
type != 1 PASS
!strncmp(ent->d_name, "ctc", 3) PASS
type != 256 PASS
got PASS && PASS && PASS = PASS
+ if (type != 1 || (!strncmp(ent->d_name, "ctc", 3)&&
type != 256))
Here the last two expressions are and'ed and that result is or'ed with
the
first expression.
Test 1:
type = 1
ent->d_name = "foo"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) FAIL
&& FAIL
type != 256 PASS /
got FAIL || FAIL = FAIL
Test 2:
type = 1
end->d_name = "ctc"
type != 1 FAIL
!strncmp(ent->d_name, "ctc", 3) PASS
&& PASS
type != 256 PASS /
got FAIL || PASS = PASS
Test 3:
type = 2
end->d_name = "ctc"
type != 1 PASS
!strncmp(ent->d_name, "ctc", 3) PASS
&& PASS
type != 256 PASS /
got PASS || PASS = PASS
continue;
new = calloc(1, sizeof(struct device));
I feel confused, but shouldn't the condition stand:
if ( type != 1 && !(!strncmp(ent->d_name, "ctc", 3) && type == 256) )
That is: ignore the same as before except for device that has type 256
and starts with "cnc"?
(original condition was: if (type != 1))
We went from FAIL, FAIL, PASS to FAIL, PASS, PASS with the addition of
the
parens. Without the patch, the code requires ctc devices to be type
!= 1 and
type != 256, when we really need to just check that type != 256 if we
have a
ctc device.
Hm, the comment in the original patch says:
/* S390 channel-to-channnel devices have type 256 */
so I think your second case is not so relevant (my proposition
would FAIL, FAIL, PASS), but following case is:
(A)
type = 256
end->d_name = "ctc"
and the goal of the original patch and the BZ (as I understood it)
was not to ignore this device - therefore FAIL.
I assume both name and type are important, so this:
(B)
type = 256
end->d_name = "foo"
should be ignored - PASS
This patch came up as a fix of the problem that
with original patch we stopped to ignore lo device:
(C)
type != 256 (don't know what exactly)
end->d_name = "lo"
with
- if (type != 1&& !strncmp(ent->d_name, "ctc", 3)&& type !=
256)
FAILS, but with
+ if (type != 1 || (!strncmp(ent->d_name, "ctc", 3)&& type
!= 256))
PASSes (that is, lo device is ignored as before).
To sum up:
expected behavior - as I understand it:
A: FAIL
B: PASS
C: PASS
- if (type != 1&& !strncmp(ent->d_name, "ctc", 3)&& type !=
256)
gives :
A: FAIL
B: FAIL
C: FAIL
+ if (type != 1 || (!strncmp(ent->d_name, "ctc", 3)&& type
!= 256)):
gives:
A: PASS
B: PASS
C: PASS
my suggestion
if ( type != 1 && !(!strncmp(ent->d_name, "ctc", 3) && type == 256) )
gives:
A: FAIL
B: PASS
C: PASS
Radek
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
11-04-2010, 02:53 PM
"Brian C. Lane"
Fix test for CTC devices from yesterday.
On Thu, Nov 04, 2010 at 01:34:29PM +0100, Radek Vykydal wrote:
> my suggestion
> if ( type != 1 && !(!strncmp(ent->d_name, "ctc", 3) && type == 256) )
> gives:
> A: FAIL
> B: PASS
> C: PASS
I agree with Radek here. This is how I examine these kinds of problems:
type, name 1 !(2&&3) 2 3 Result
1 bob F T F F F skip continue
2 sam T T F F T continue
256 sam T T T F T continue
256 ctc T F T T F skip continue
I like putting the type check first because that will short-circuit the &&
preventing needless calls to strncmp. And some may say I use () excessively,
but it prevents confusion when evaluating long expressions.
--
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
Fix test for CTC devices from yesterday.
