Re: haskell transition: Remove, Transition, Upload Fix or Upload Fix, Wait, Transition?

[Mehdi Dogguy <mehdi@dogguy.org>]

On 08/06/12 11:33, Joachim Breitner wrote:
> besides some very few remaining rebuilds on mipsel and sparc, the
> haskell transition is almost ready, if there were not a release
> critical bug in haskell-cryptocipher affecting big-endian machines
> (mips, mipsel, ppc, s390x, s390x, sparc).
>
> I have finally come up with a proper fix, which requires patching
> haskell-cpu and haskell-cryptocipher, and is on its way to
> experimental to see if it actually works :-)
>
> I see two modes of action now:
>
> A) In order to not hold up the transition: 1. remove all affected
> packages (i.e. all reverse dependencies of haskell-cryptocipher on
> these 6 architectures) in unstable 2. let the transition happen 3.
> upload the fix to unstable 4. let all these packages build again 5.
> transition them

IOW, remove the following list?

 haskell-cryptocipher
 haskell-clientsession
 haskell-cprng-aes
 haskell-tls
 haskell-tls-extra
 haskell-authenticate
 haskell-hledger-web
 haskell-http-enumerator
 haskell-yesod-core
 haskell-yesod
 haskell-yesod-auth
 haskell-yesod-form
 haskell-yesod-json
 haskell-yesod-persistent
 haskell-yesod-static

(Computed using "dak rm" so might be an over-set)

> or B) 1. Upload the fix to unstable now. 2. Wait for all reverse
> dependencies be built on these six architectures. 3. Do the
> migration then.

If the fix and other packages are ready, then what will we gain by 
removing haskell-cryptocipher (and r-deps) from testing? If they are 
really fixed in sid, I think we can age them to let haskell stack migrate.

But looking at the actual state, there is stuff not ready yet:
- haskell-tls doesn't look ready at all (but listed above)
- as a consequence, haskell-tls-extra is not ready too (but listed above)
- haskell-cprng-aes is not too (but listed above)
- haskell-diagrams is not ready
- the following packages need to go away from testing first (that
shouldn't be an issue though): haskell-happstack-util,
haskell-happstack-state, haskell-happstack-ixset, haskell-happstack-data.
- probably more…

I'll will do a britney test run later to see what is really needed to
fix/remove/age.

> I’m worried that B would make the large transition happen after the
> freeze.

We were already worried when the transition started without
coordination.

Regards,

--
Mehdi Dogguy مهدي الدڤي
http://dogguy.org/