pcp
[Top] [All Lists]

RE: qa/079 vs od(1)

To: "'Nathan Scott'" <nathans@xxxxxxxxxx>
Subject: RE: qa/079 vs od(1)
From: "Ken McDonell" <kenj@xxxxxxxxxxxxxxxx>
Date: Fri, 6 Mar 2015 17:58:57 +1100
Cc: "'PCP'" <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <933331927.21934728.1425601253534.JavaMail.zimbra@xxxxxxxxxx>
References: <1739506169.21934177.1425601023497.JavaMail.zimbra@xxxxxxxxxx> <933331927.21934728.1425601253534.JavaMail.zimbra@xxxxxxxxxx>
Thread-index: AQJQ+pv1cMq41VCV9et4piOLE+wgX5wNv90g

> -----Original Message-----
> From: Nathan Scott [mailto:nathans@xxxxxxxxxx]
> Sent: Friday, 6 March 2015 11:21 AM
> To: Ken McDonell
> Cc: PCP
> Subject: qa/079 vs od(1)
> 
> Hi Ken,
> 
> Test 079 is failing here for a recent OS update, looks like a change in
> behaviour for od(1).  Comments in the test suggest this kind of thing may
> have happened before (hence the use of tr).
> 
> :: [  BEGIN   ] :: Running 'diff -urN /var/lib/pcp/testsuite/079.out
> /var/lib/pcp/testsuite/079.out.bad'
> --- /var/lib/pcp/testsuite/079.out    2015-03-02 19:59:06.000000000 -
> 0500
> +++ /var/lib/pcp/testsuite/079.out.bad        2015-03-04 07:44:38.432624261 -
> 0500
> @@ -1,7 +1,7 @@
>  QA output created by 079
>  Dump first 128 bytes ...
> -0000000 \0  \0  \0 004   P 005   & 002  \0  \0   ]   |   7   }   q   #
> -0000020 \0  \f 022   :  \0  \0  \0  \0   m   o   o   m   b   a  \0  \0
> +0000000 \0  \0  \0 204   P 005   & 002  \0  \0   ]   |   7 375   q   #
> +0000020 \0  \f 022 272  \0  \0  \0  \0   m   o   o   m   b   a  \0  \0
>  0000040 \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0
>  *
>  0000120 \0  \0  \0  \0  \0  \0  \0  \0   E   S   T   -   1   1   E   S
> 
> 
> Any reason we should not switch to using the -a option to od here?
> The man page states that ignores the high bit issue, and indeed it seems to
> produce reliable output once more.  Since this is simply dumping the start of
> a fixed log at the start of the test (I guess for sanity checking?) this 
> seems a
> valid approach... thoughts?  Or, should we put this initial output into .full
> perhaps, since its not exercising any PCP tools?

I think the data is actual useful ... if it ever changes the following PCP test 
is likely to fail.

Probably better to skip the binary parts of the header and label record and 
report the bits that matter, just the hostname and TZ ... neither of which 
contains high-order bits.

I have this done now, and will push in my next batch for others to review.

<Prev in Thread] Current Thread [Next in Thread>