Tim Rightnour <root@garbled.net> writes:
>On this note, I'd like to say.. that changes to sysinst, that need to be pulled
>up.. should be verified on at *least* 2-3 platforms.  Sysinst is the one part
>of the system we absolutely cannot afford to screw up.  Thats our first line of
>contact with the user.

I agree.  I'm doing the majority of my work and testing here on the an
x86, but before i'm done i plan to try it on other archs, and i
_strongly_ encourage others to do the same.


>I would even go as far to say that releng should make sysinst a mandatory
>testing item before a release, and that serious bugs should delay any release
>cycle.

"You must test it!  If you don't test it, we'll punish you!"

"You'll punish me?  How?"

"We'll send you lots of annoying mail!"

"So, what you're saying is, i should invest now in a smart MUA?"

"But you're not supposed to ignore us, we're important!"

"But, what does it get me to listen to you?"

...

Hell, half the time it seems like the only people who actually give a
damn about making releases _at all_ are releng.  (That's not really
true, but there's been way too much "who needs releases?" spread
around by developers to discount it.)

Given the level of, uh, commitment to detail and testing the sysinst
has seen in the past, I don't see any reason to believe that it'll get
any more seriously tested if releng asks than if it doesn't.


>This is all IMHO of course.. but I think the SIGFPE stuff in 1.4 is a complete
>fiasco, and makes us look terrible.  I'd also like to publicly thank Chris for
>taking the time to fix that, and other important glitches.

Well, I didn't fix it.

I've currently figured out what I _think_ is going wrong, but i'm next
to clueless about PC partitioning issues so I don't want to make a
mess of it.  I put my analysis in my PR on the subject (7608), and
i've poked fvdl a couple of times to look at it.  He's apparently
rather busy at work right now, but hopes to look at it within a week.
(*sigh*)


As for the rest, well, you're welcome.  To be quite honest, the reason
why I did it is because the inconsistencies and brokenness i saw
boggled my mind, and made me sufficiently annoyed at the program to go
in and fix them.  It amazes me that we (the project) had the chutzpah
to ship this to users as an installation tool...


cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.