rawhide report: 20060420 changes

[Richard Hally]

Jeff Spaleta wrote:
> On 4/20/06, David Nielsen <david at lovesunix.net> wrote:
>> Those packages were broken all through the FC5 cycle, clearly who ever
>> owns them doesn't care about them enough to keep them in shape for
>> testing.
>>
>> Could we please nuke them?
> 
> I'm pretty sure those packages were rebuilt for the test iso trees. 
> The reality is without automated infrastructure which triggers kernel
> module rebuilds in a timely manner any kernel module packages will
> have an extremely difficult time tracking rawhide, which can see daily
> kernel changes.
> 
> This is not a failing due to a maintainer's lack of "caring." Addon
> kernel modules in Core, no matter who maintains them, are going to
> need scripted infrastructure which triggers a rebuild in the build
> system as soon as a new kernel is built.
> 
> You simply can not expect a human being to be able to respond to a
> kernel build notification at a random time in the middle of the
> night...every night... so the gfs kernel module packages can be built
> in time for the daily rawhide push to the public servers.   You have
> to remember rawhide is designed to gets pushed once a day, a kernel
> module maintainer can't show up after the tree goes public in the
> morning and request a build and have it pushed to the public trees
> later that day. The strict dependancy on kernel version/release that
> kernel module packages must meet is unreasonably difficult to track
> via human effort on a daily basis given the way rawhide releases are
> structured now.  It can be done when the rawhide tree enters a period
> of freeze however, which is exactly what happens when the test iso
> snapshots are spun up.
> 
> -jef
> 
since they have "special" requirements, perhaps they could be separated 
into an other repo e.g. devel-modules (or devel-special)?