pcp
[Top] [All Lists]

RE: QA temporaries

To: "'Nathan Scott'" <nathans@xxxxxxxxxx>
Subject: RE: QA temporaries
From: "Ken McDonell" <kenj@xxxxxxxxxxxxxxxx>
Date: Thu, 28 Nov 2013 12:49:42 +1100
Cc: "'PCP'" <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <632473151.19046786.1385590121642.JavaMail.root@xxxxxxxxxx>
References: <1626967316.19041413.1385589559423.JavaMail.root@xxxxxxxxxx> <632473151.19046786.1385590121642.JavaMail.root@xxxxxxxxxx>
Thread-index: AQIHF8tGpGBS5VqrImQGasOgsLOPn5nJh4Eg
Good idea ... $seq-$$ might be better just in case some bogan script is 
expecting $tmp.foo to expand to something with a single dot.

We should extend this to adding "dodgey version from QA/$seq" comments to all 
the configuration files we replace during QA, for much the same reason.

The only technique I've found useful is ls -lt on the temp file and ls -ltu on 
the qa/[0-9]*[0-9] files (assuming you've not done a grep there in desperation 
in the recent past).

-----Original Message-----
From: Nathan Scott [mailto:nathans@xxxxxxxxxx] 
Sent: Thursday, 28 November 2013 9:09 AM
To: Ken McDonell
Cc: PCP
Subject: QA temporaries

Hi,

I'm finding quite a few tests fail to cleanup temporaries and its fairly 
painful to try work out which ones after a full QA run.  As a result, I'm 
thinking of tagging temporaries with a test sequence number for diagnosis, 
e.g....

qa$ diff common.rc common.rc.newtmp
85a86
> [ "X$seq" != X ] && tmp=/tmp/$seq.$$

There is test output filtering work to be done as a result of
this kind of change, but I need to do something as my /tmp is
starting to look like a dogs breakfast.

Is there a better way you know of to trace these files back to
the culprit test?

cheers.

--
Nathan


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