[pkg-boost-devel] [VAC] now - undefined (was Re: ICU transition status)

[Jay Berkenbilt]

Luk Claes <luk at debian.org> wrote:

> Jay Berkenbilt wrote:
>> Domenico Andreoli <cavok at dandreoli.com> wrote:
>> 
>>> Unfortunately I am having connectivity problems, I am changing DSL
>>> provider and (almost) anybody in Italy knows how much painful is such
>>> a change.
>>>
>>> Please take care of boost wrt ICU and any other important issue it
>>> might have.
>>>
>>> Best regards,
>>> Domenico
>> 
>> If I have time before my vacation and no one from the debian boost
>> team responds, I can NMU this.  I would just take 1.34.1-3 (currently
>> in experimental) and update the changelog and control as needed.  I've
>> done this locally and am building in pbuilder.
>> 
>> I seem to recall an earlier discussion that the only reason 1.34.1-3
>> was uploaded to experimental instead of unstable was to avoid
>> lengthening the boost 1.33 to 1.34 transition, so that's why I'd base
>> the NMU off of 1.34.1-3 rather than 1.34.1-2.
>
> Well, you may recall correctly, though things have changed in the
> meantime... There is not supposed to be an upload of boost that isn't
> backwards compatible!

I hadn't looked carefully at the packages yet.  Now that I am looking
at it, I see clearly from the changelog that 1.34.1-3 was an ABI
change, so you're right -- I definitely don't want to upload it.
(Though I'm surprised that going from gcc 4.1 to 4.2 is really an ABI
change.)

>> I'm getting ready to go offline for three weeks myself, so it's not
>> the best time to do an NMU, but this one is sufficiently trivial that
>> it should be pretty safe.
>
> It's not as trivial as you seem to think, so please refrain from
> uploading it.

Point taken.  I would feel the same way if someone were going to NMU
tiff or ICU.

In any case, I wouldn't have uploaded without testing carefully.  I
have software that uses some of the boost libraries.  I would at least
have locally installed boost and checked my software, openoffice, and
perhaps some other reverse dependencies.  My initial comments were
based on my memory of earlier conversations, not a careful analysis.
But you're right, this is not to be taken lightly.

In any case, I can either do an NMU based on 1.34.1-2 (which really
should be safe since 1.34.1-2 is already in testing), or I can just
drop it and let someone else take care of it.  Unless someone says to
go ahead with 1.34.1-2.1, I'll just leave it alone.  If it is not
resolved by the time I'm back from vacation, I probably will upload
1.34.1-2.1 though.  Given that this fixes a 10-day-old RC bug, anyone
could do an NMU at this point anyway.

--Jay