On Mon, May 18, 2009 at 10:49 PM, Zach Welch <[email protected]> wrote:
> Hi all,
>
> Since this weekend, I have spent some time working on a set of Perl
> scripts to prototype the back-end system that will aggregate the data
> and present it for review (pre-analysis).
>
> I started on this to replace the static developer table that I was using
> to augment The List. I think most of what I am doing will need to be
> rewritten in SQL to scale up, but I wanted the first draft to avoid any
> serious external dependencies (e.g. a SQL engine).
>
> These scripts work on plain text table files: one line per row, '\t'
> delimits the columns. The 'tool.pl' script can be used to manipulate
> the data from the shell (or other non-Perl scripts), and the 'html.pl'
> script can convert the data into cross-linked static HTML pages. Both
> use the included 'Tables.pm', which I whipped together this weekend for
> this purpose.
>
> The tables track reports from users detailing the results from running
> the regression test suite. A "report" will specify the "platform",
> "interface", and "target" or "board" that was tested by the "user",
> along with the 'version' of OpenOCD, the test 'results', along with a
> 'comment'. In this description, words in double-quotes ('"')
> effectively describe various primary tables, while the words in
> single-quotes ("'") describe column fields.
>
> Similarly, each user may specify a list of devices that they can use for
> testing, so specific users may be prompted for testing particular
> configurations. This last feature also allows me to create a developer
> roster that improves on the one that I had been tracking in The List
> (until r1828).
>
> Presently, this user-tracking half is (all but) complete, since that was
> the reason I started this work. The reporting then builds on it; users
> with one platform, interface and target should not be made to specify
> these parameters repeatedly when they will be contributing a single
> time-series of data points.
>
> Known problems:
> - I have tried to KISS, so there are going to be definite limits.
> - Additional database design work needs to be done.
> - Static output only.
> - Missing an 'email.pl' script to provide a gateway from e-mail messages
> sent by the automated tester.
> - Automated tester does not report PASS/FAIL and send e-mails.
>
> I will try to post my initial implementation in the next day or two, but
> I am curious what others think of this approach.
>
> Cheers,
>
> Zach Welch
> Corvallis, OR
>
> _______________________________________________
> Openocd-development mailing list
> [email protected]
> https://lists.berlios.de/mailman/listinfo/openocd-development
>
I think that would be great. As a user what is our interface to this? Do we
just send an email with lines in the specified format? I haven't used the
regression test suite so maybe that would explain this some.
One way to aggregate the data would be to create a web form that users could
fill out manually, but also have another perl script in the test suite that
could automatically fill out the form. The most elegant way would be to tie
the form directly to the SQL database, but for something quick & dirty we
could just use a google form like this: (
https://spreadsheets.google.com/viewform?formkey=cnpOSWRCcE9DRjl1a3E5NURnUFQ4Rnc6MA..),
and then pull a csv file of the data when we wanted to. Just a thought. I
know some PERL & SQL so I'm willing to help out.
thanks,
Paul
_______________________________________________
Openocd-development mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/openocd-development
