Re: [pkg-boost-devel] Bug#454605: Boost & gcc 4.1 (was Bug#454605: build against new icu 3.8 packages)
Steve M. Robbins wrote:
> On Sun, Dec 09, 2007 at 03:55:31PM +0100, Pierre Habouzit wrote:
>> On Sun, Dec 09, 2007 at 01:44:05PM +0000, Steve M. Robbins wrote:
>>> On Thu, Dec 06, 2007 at 05:55:29PM +0100, Domenico Andreoli wrote:
>>>> Hi,
>>>>
>>>> rebuilding Boost as-is would require a new Boost transition. Could
>>>> we continue to build it with gcc 4.1? At least until 1.35.0 is out,
>>>> that could be a few months away.
>>> On the other hand, gcc default is version 4.2 since September 1. It
>>> would be nice to have boost built with the default gcc.
>>>
>>> Couldn't we do the transition now?
>> I'm sure the RMs are thrilled to go for a new 2 to 3 months long
>> migration. Couldn't boost be less insane and stop bumping sonames for no
>> good reasons instead ? I mean what's the point of depending on the gcc
>> or glibc version, whereas Debian already tracks that down for them.
>
> Point taken about encoding gcc version in the SONAME. Mind you that
> would still imply another transition! :-)
>
> I don't understand, however, why such a transition is so onerous.
>
> For example, why doesn't uploading a library with a new SONAME simply
> cause the depending packages (identified by the algorithm that
> generates "excuses") to be scheduled automatically for a rebuild?
Because you forget about maintainers uploading packages in the meantime
that break for some reason or another or that tangle the transition with
other unrelated transitions and that it's not enough to schedule
rebuilds, the rebuilds all have to succeed and been uploaded...
Cheers
Luk
Reply to: