On Mon, May 14, 2018 at 07:13:57PM +0200, Emilio Pozuelo Monfort wrote:
> On 13/05/18 16:08, Adrian Bunk wrote:
> > Control: reassign 898465 src:icu 60.1-1
> > Control: retitle 898369 boost: ICU version used is part of the ABI
> > Control: retitle 898465 ICU must not migrate to testing before the boost
> > ABI breakage is resolved
> > Control: affects 898369 libmapnik3.0 viking
> > Control: block 898465 by 898369
> > Control: block 894159 by 898369 898465
> >
> > On Thu, May 10, 2018 at 06:06:08PM -0300, Thadeu Lima de Souza Cascardo
> > wrote:
> >> Source: boost1.62
> >> Version: 1.62.0+dfsg-5+b2
> >> Severity: serious
> >>
> >> After upgrading boost1.62 to 1.62.0+dfsg-5+b2, ncmpcpp does not start
> >> anymore.
> >>
> >> $ ncmpcpp
> >> ncmpcpp: symbol lookup error: ncmpcpp: undefined symbol:
> >> _ZNK5boost16re_detail_10620031icu_regex_traits_implementation12do_transformEPKiS3_PKN6icu_578CollatorE
> >> ...
> >
> > boost::re_detail_106200::icu_regex_traits_implementation::do_transform(int
> > const*, int const*, icu_57::Collator const*) const
> >
> > Yes, boost makes the ICU version it uses part of its ABI.
> >
> > There is no obvious good way forward, the RC bug in ICU will ensure that
> > the breakage won't migrate to testing before this got sorted out.
>
> boost1.62 should:
> - bump the libicu-dev build-dependency to ensure it builds against icu 60
> - add breaks against the rdeps that use the affected symbols
>...
For the opposite problem this also requires:
- boost1.62 to add versions to the shlibs for the affected libraries
- binNMUs of all users of these symbols to force a dependency on the
icu 60 boost
> Cheers,
> Emilio
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
