[forum] Some perspective from the cheap seats...

[Stuart Anderson]

On Fri, 21 Mar 2003, Alan Cox wrote:

> Providing the tree has regular "most of it compiles" snapshots and
> developers and others are encouraged to try them out that isn't as true
> as you think. In fact XFree has singularly failed to achieve this
> goal because its not possible.

There are lots of things that can be said about this.

1) The cvs is accessible. We effectively have daily snapshot available.

2) A kind of catch-22 has evolved here. People don't both building daily
   CVS because they expect it to be broken. Because no one bothers building
   daily, breakage is not detected, or reported. I've seen several reports
   of build breakage in the final days of 4.3, and almost always, there was
   a cvs message containing the commit to fix the problem sitting in my
   mailfolder folder already (ie there was a quick turnaround).

3) I would like to see a slightly more formal production of snapshots
   happening, and have talk with a couple of people about doing this. It
   won't get done overnight, but it has been done before, and can be done
   again.

   One complication here is that there are so many different platfoms for
   which regular build should be performed. Also, a lot of people are probably
   more comfortable install packages provided by their OS vendor. These are
   both issues that will need to be taken into consideration when planning
   a mor formal snapshot system.

> On 4.3 my Riva128 failed and my Glint
> Delta no longer works. I'd bet no XFree core member even has a Glint
> Delta on PCI let alone tested it on 4.3. They have better things to be
> doing anyway.

Please make sure you file a bug at http://bugs.xfree86.org/ so this will be
properly tracked.

> A test suite that uses each XAA op in various ways would help no end, as
> would one which ran through lots of mode settings.

There is a formal test suite for X. It cycles through all of the possible
combinations for every operation, at the API/Protocol level. This should map
to good coverage of XAA even though the tests weren't written explicitely
for that.

The three difficulties with this are 1) The test is very large, and takes
a long time to run, 2) the test is non-trivial to setup, thus discouraging
people from using it, and 3) the combination of cards and modes/resolution
and depths simple explodes. You could keep a large farm of systems busy 7x24
just running through all of the possible combinations.

The LSB has taken similar test suites, and packages them up and added enough
stuff to them to nearly automate their use. Doing this for either the XTS or
VSW4 is in the queue. Hopefully, this can address the 2nd difficulty, and
possibly indirerctly reduce the other 2 as well. This package will only be
useful on Linux, but fortunately, since the graphics drivers are OS
independent anyway this won't be an issue.


                                Stuart

Stuart R. Anderson                               anderson@netsweng.com
Network & Software Engineering                   http://www.netsweng.com/
1024D/37A79149:                                  0791 D3B8 9A4C 2CDC A31F
                                                 BD03 0A62 E534 37A7 9149