stripping off "xf86-*-" from drivers

David Miller davem at davemloft.net
Sun Jan 20 16:19:09 PST 2008


From: Luc Verhaegen <libv at skynet.be>
Date: Sun, 20 Jan 2008 22:59:24 +0100

> Why can anyone agree with "Let's go monolithic again so that i 
> still do not have to bother in any way about being slightly backwards 
> compatible".

The whole point is that this backwards compatibility is not happening
at all, the dependencies are more complex then before, and as a result
developers make changes that break the build of half the drivers.

We were definitely better off before.

> I don't think there has been anything worse than this.

The GC private changes for SELINUX were just as bad as the
libpciaccess case, the input driver API changes are another prime
example of how this modular stuff simply does not work.

> This is one part of the solution. Tinderbox should be revived, and those 
> who commit stuff to master or a release branch need to get blamed for 
> any build breakage. This will solve a lot of the issues already. A lot, 
> but not all.

How about getting developers ACTUALLY TYPING "make"!  That is all the
is being asked of them.

If you need Tinderbox to enforce this, the necessary developer mind
set is simply not there.  If developers think it's OK to not have
to test the build, Tinderbox doesn't change that.

With a non-modular tree we could at least reasonably enforce getting
developers to test the build of any API changing patches with a
"everything enabled" build.

Anyone with write access must be willing to do this.

You have to force the hand of people when it gets to the point
it is right now.

Here is one fact I find disturbing.  The person who left several
input drivers non-buildable for more than a year still checks
code into the tree, and still defends his actions in one way
or another every time it is brought up.

Nobody who has any power wants to reprimand him or others who
do things like that.  It is proof that the mentality necessary
to handle the modular tree properly is simply not there, and
therefore your suggested "fix" is pure fantasy.



More information about the xorg mailing list