[plt-dev] DrDr Notifications

From: Jay McCarthy (jay.mccarthy at gmail.com)
Date: Fri Feb 12 15:09:16 EST 2010

I've just updated DrDr so it will send notifications when a file's
status may be worth investigating. Statuses worth investigating are
considered: timeout, unclean exit, and stderr output. Output changes
are NOT considered worth investigating for now.

--

I've updated the DrDr UI as well. Look at: http://drdr.plt-scheme.org/18066/

Underneath the commit information there is a status summary organized
by "responsible" party with links to the files.

Whenever there is a change to the output, such as on

http://drdr.plt-scheme.org/18066/collects/tests/mzscheme/filelib.ss

There is a link to a page that shows the difference:

http://drdr.plt-scheme.org/diff/18066/18064/collects/tests/mzscheme/filelib.ss

--

Output changes are the biggest problem right now for DrDr.

Some files do random tests:

http://drdr.plt-scheme.org/diff/18066/18064/collects/tests/net/main.ss

Some refer to time:

http://drdr.plt-scheme.org/diff/18066/18064/planet/cce/scheme.plt/4/1
http://drdr.plt-scheme.org/diff/18066/18064/collects/tests/srfi/19/tests.ss

Some refer to temporary files (that are random):

http://drdr.plt-scheme.org/diff/18066/18064/collects/tests/mzscheme/filelib.ss

Some print performance information in a way I don't parse yet:

http://drdr.plt-scheme.org/diff/18066/18064/collects/tests/mzscheme/benchmarks/common/auto.ss

And some change for reasons I can't understand:

http://drdr.plt-scheme.org/diff/18066/18064/src/build/make

--

Some ideas to solve this:

1. Do not notify on changes forever

2. Change the files to not change (or make a cmdline option that turns it off)

3. Write wrapper programs just for DrDr in tests/drdr that filter the
changing text

4. Add an SVN property to control what DrDr ignores

--

I don't like 1 because I don't think 30-ish files should prevent a
useful piece of data from being used for the thousands of other files.

2 isn't satisfying because the changing information may be useful for
humans and some changing may be necessary (printing of random seeds
and random programs to debug when something breaks.)

I like 3 because it is flexible, but it makes DrDr a deeper part of
the tree than it already has become.

Eli likes 4 because it keeps DrDr more independent. I don't like 4
because I want to "keep everything in the language". I can see the
property becoming equal to 3 except through an awkward interface.

--

Any thoughts?

Jay

-- 
Jay McCarthy <jay at cs.byu.edu>
Assistant Professor / Brigham Young University
http://teammccarthy.org/jay

"The glory of God is Intelligence" - D&C 93


Posted on the dev mailing list.