[Pkg-scicomp-devel] scotch and libmesh

[Ondrej Certik]

On Sat, Aug 9, 2008 at 12:26 AM, Adam C Powell IV <hazelsct at debian.org> wrote:
> On Thu, 2008-08-07 at 17:34 +0200, Luk Claes wrote:
>> Adam C Powell IV wrote:
>> > On Wed, 2008-08-06 at 17:57 +0200, Luk Claes wrote:
>> >> On Mon, Aug 04, 2008 at 12:23:53PM -0400, Adam C Powell IV wrote:
>> >>> On Thu, 2008-07-31 at 20:08 +0200, Luk Claes wrote:
>> >>>> Adam C Powell IV wrote:
>> >>>>> Greetings,
>> >>>>>
>> >>>>> I am writing to request that scotch 5.0.6-1 (uploaded 7/21) and libmesh
>> >>>>> 0.6.2-2 (uploaded 7/25) be included in the Lenny release.  Scotch is in
>> >>>>> the NEW queue because this release added shared libraries.  The libmesh
>> >>>>> release closes a lot of bugs, and uses the scotch shared libraries, so
>> >>>>> it is in Dep-Wait state waiting for the new scotch package.
>> >>>> libmesh unblocked, scotch can only be processed once it's accepted in
>> >>>> unstable: so no decision on scotch yet. Please poke us again once it has
>> >>>> been accepted in unstable.
>> >>> *Poke* (Scotch accepted, is in unstable, but marked as frozen on
>> >>> bjorn.haxx.se).
>> >>  257 files changed, 10405 insertions(+), 5707 deletions(-)
>> >>
>> >> I'm afraid this is too much to review, so I'm not going to unblock it...
>> >
>> > Wow, even though I uploaded it six days before the freeze?  Sorry to
>> > hear that.  In my past experience, new version upgrades have taken 3-4
>> > days to get through NEW, not two weeks...
>>
>> Well, the freeze month was announced in february, not just a couple of
>> weeks before it happened...
>
> In Marc's February message: "Please don't wait with uploads for the last
> day before the freeze, thanks," which I didn't do.  (Most of my other
> packages were release-ready by early June.)  And I read the message from
> Adeodato Simó on July 22 as "Don't panic, if you upload before the
> freeze you'll get in".
>
> Sounds like the difference is between "uploaded" and "in unstable" based
> on inconsistency between the two messages.  My mistake, but an honest
> one.  Today is still not that long after the freeze, is it not
> reasonable to have an exception?
>
>> > Why isn't getting through the NEW queue sufficient review?
>>
>> Because the NEW queue check is to check if there is nothing wrong
>> regarding FTP Master and legal rules to have the package in the archive
>> (unstable).
>>
>> A freeze exception is to get a package from unstable to testing...
>
> Right, but what's the difference in quality standards?  I guess I should
> RTFM...  Again, seems a little arbitrary given the upload date.
>
>> > If this decision stands, I'll need to re-upload libmesh to change its
>> > scotch dependency in order to fix this.
>>
>> Why, do the packages in testing not work?
>
> The new libmesh closes seven bugs, including broken examples, support
> for I/O in only one format which requires a non-free visualization
> system (GMV, which is not in Debian), and missing links to scotch libs
> -- which would require scotch shared libs to fix.  The new scotch closes
> three bugs, but one is "new upstream" so that doesn't count, another is
> lintian fixes, and the third is shared libs.
>
> So what should I do now?
>     1. Do nothing, let all the bugs in testing enter the lenny release
>     2. Do nothing, get a freeze exception for the new scotch, which
>        would let libmesh go in
>     3. Upload scotch 5.0.1.dfsg-1lenny1 (which needs to go through the
>        NEW queue, but the changes are much smaller than 5.0.6) and
>        libmesh 0.6.2.dfsg-1lenny1 depending on it
>     4. Upload libmesh 0.6.2.dfsg-1lenny1 depending on the old scotch
>        with only static libs, but closing the other six bugs in libmesh

I think Luk is right, that this is quite a big change, which we should
have uploaded much before than 6 days before the freeze.

On the other hand a broken package in Lenny is a shame, so anything
from the above 1.-4. should happen. If you ping me, I'll help you test
the package in case of 3. or 4..

Ondrej