Mike A. Harris wrote:
s/fixed/attempted to fix/..... the problem is still present.
2 people have suggested using rpm triggers to solve this problem.
Triggers are a risky business on their own however which can end
up creating problems days/weeks/months down the road that are
completely impossible to cleanly fix. Not that a trigger can't
be done correctly the first time and not have any problems, but
rather that experience of my own, and that of others has shown
that triggers are very tricky and statistically are not bug free
the first time around, leading to bigger problems that only
manifest in future upgrades, and aren't bypassable automatically.
As such, I'd like to avoid adding triggers at all costs.
Altogether, this leaves 3 alternate scenarios that I can
think of:
1) Let anaconda or something else fix it during the OS
upgrade cycle, via voodoo magic.
2) Use rpm triggers with no guarantee of it actually fixing it,
and no way to 100% predict the future, along with all of the
associated risks of using triggers.
Downside to #1 is that users upgrading manually using rpm -Uvh,
or via up2date, yum, apt will have a one time growing pain during
the transition from XFree86 to xorg-x11. Solution #1 is what we
probably would have done in any previous OS release to play
things safe.
Downside to #2 is the risks involved with triggers, that have
shown again and again repeatedly in almost every rpm package that
has ever used them, that triggers are very hard to get right, and
to predict all the possible ways the script might get called in
the future. They're notoriously hard to test in advance also,
but once they're out in the wild, if a bug is found, then users
are essentially screwed until they've upgraded at least 2 times.
Again, I'm very hesitant to use rpm triggers, but at this point
nothing is ruled out yet. I'm open to suggestions.
#1 need not be limited to anaconda. It could also be a super special
case that apt, yum and up2date handle because it is extremely rare but
important.
#2 triggers MAY not be a bad thing if the implemntation is good, and
unlikely to introduce any security related problem. What specific
concerns do you have about triggers in this case?
Warren