Using TBoxes on Branches - 11/15/07 Meeting Notes

Present: John, Aparna, Dan, Bryan, Bear, Heikki, Jared, Mikeal

Issues

• John tried to get his branch to run on TBoxes as mentioned in our new process^? . It didn't work. John wasted lots of time trying to make it work and had to fight lots of bugs, most not related to his branch, before being successful (and only partially).
• Take home messages from John:
  • All the issues found turned out being real issues tracked back to Chandler bugs. So, in that sense, that was good but concerning we didn't find those issues any other way.
  • It's hard to track issues when they happen on TBoxes
  • There's no patterns in those bugs unfortunately. We had: TBoxes use of "make distrib" instead of "make install", merge problems, makefile bugs

Solutions

• We need a way to make debug on TBoxes easier: use Wing remote debugger
• Try to run part of the tests on the TBoxes to accelerate the bug cycling
• Try to run TBoxes locally i.e. on developers machines
• Have more TBoxes in more various environments

Actions

• Bear: Create a set of tools/docs to explain to devs how to run TBoxes on their machines so that they can run that all night for instance before merge branches
• Philippe: Change the process page to take that into account
• Heikki: TBoxes should take arguments to run only a subset of tests
• Stearns: Write a doc explaining how to remote debug with Wing
• Heikki: Have a good doc on the wiki describing all the parameters of the TBoxes (OS, RAM, Disk, Time Zone used, Locale, etc...)

Recorded_script branch

That was a session to talk specifically about John's branch landing:

• Doesn't work on Linux: Bear working on it, has to do with the fact that recorded_script requires the UI
• Decided to land recorded_scripts early in the 0.7.4 cycle (too many things to polish to merge it in 0.7.3 trunk right now, too risky)
• Dan working off the branch in the next few weeks to do his script tasks and test some more

-- PhilippeBossut - 16 Nov 2007