[AWFFULL] no results some months

[Steve McInerney]

On Fri, 2010-05-14 at 10:10 -0400, Zack Galbreath wrote:
> Hi,
> 
> First off, thanks for making this tool.  We use awffull primarily to 
> estimate how frequently our software tools are downloaded in any given 
> month.  The problem we're experiencing is that some months awffull 
> reports almost no traffic to our site.  You can take a look for yourself 
> here:
> 
> http://public.kitware.com/stats/awffull/
> 
> 
> Do you have any idea why this might be happening?  Can you suggest any 

That looks like something really bad is happening. Which I kinda guess
you'd already realised. :-)


> places where we can look to help diagnose this problem?  What would I 

Do a manual run of just the given month when you get no details, for
starters. Or perhaps better - sneak up on it:
Do the 1st of April 2010; then run for the 2nd and so on.
After each day; check if data is being added and displaying correctly.
Obviously, when you do this testing - do it manually and set the output
directory elsewhere so you don't clobber the production display.

It looks like *something* is causing the entire months data to get wiped
out. I have a suspicion as to what may cause this, but ... i thought
that bug long squashed. Ahh. v3.8.2. very old indeed.
"The latest Production Release is v3.10.2, released on the 13th of
December 2008"
Hrm. I *have* been a slack bugger lately...


In essence, it's some data or order of data that awffull is properly
accounting for and hence going kersplatter.


> have to change in awffull.current if I wanted to "roll back time" and 
> regenerate the missing stats for April 2010?

Yeah, it's fiddly, but not hard to 'patch' them in as it were.
You don't need to worry about awffull.current; rather awffull.hist.

The data in that file, is what fills the front page - if the backing
page, (ie the page for April 2010) is fine, adding the numbers is easy.
Next successful run of awffull over those logs/config will auto
re-update the front page.

preserve.c is the file that controls this particular file:
order being:
month, year, hit, file, site, xfer, first_day, last_day,
page, visit

via this structure:
struct history {
    int year;                                   /* Year                                         */
    int month;                                  /* Which Month              */
    unsigned long hit;                          /* Number of Hits           */
    unsigned long file;                         /* Number of Files          */
    unsigned long site;                         /* Number of Sites          */
    double xfer;                                /* Volume Transferred       */
    unsigned long page;                         /* Number of Pages          */
    unsigned long visit;                        /* Number of Visits         */
    int first_day;                              /* First Day                */
    int last_day;                               /* Last Day                 */
};

First and Last day are, integers. ie. started on the 1st of April, last
day is the 30th of same.
These 2 are to ... *help* ensure a more accurate daily average as the
month grows.