[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#751780: marked as done (transition: libstdc++6 unordered_{map,set} ABI changes between GCC 4.8/4.9 in c++11 mode)

Your message dated Sat, 25 Apr 2015 11:27:15 +0200
with message-id <20150425092715.GN16466@betterave.cristau.org>
and subject line Re: Bug#751780: transition: libstdc++6 unordered_{map,set} ABI changes between GCC 4.8/4.9 in c++11 mode
has caused the Debian Bug report #751780,
regarding transition: libstdc++6 unordered_{map,set} ABI changes between GCC 4.8/4.9 in c++11 mode
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@bugs.debian.org
immediately.)


-- 
751780: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=751780
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems
--- Begin Message --- 
Package: release.debian.org
Severity: important
Tags: sid jessie

This change mentioned in the GCC 4.9 news:

  Improved support for C++11, including:
    The associative containers in <map> and <set> and the
    unordered associative containers in <unordered_map> and
    <unordered_set> meet the allocator-aware container
    requirements

is an ABI incompatible change in the unordered_{map,set} classes built in c++11
mode.  The libstdc++6 library itself is not affected, so instead of having a
libstdc++6 transition, all packages providing a library built in c++11 mode and
using the unordered_{map,set} classes should be transitioned (changing the name
of the library package).

A grep of all sources for #include *<unordered_{map,set}>, and then searching in
the log files for usage of -std=c++{0x,11} did show the following packages
building library packages (for packages not having verbose build logs I had to
look into the source package):

  bobcat 3.22.00-1
  c++-annotations 9.9.1-2
  capnproto 0.4.0-1
  condor 8.0.6~dfsg.1-1
  liblo 0.28-3
  mrpt 1:1.0.2-1
  qtbase-opensource-src-gles 5.2.1+dfsg-1
  rapicorn 13.07.0~ds0-1
  shogun 3.2.0-2
  tesseract 3.03.03-1
  zeroc-ice 3.5.1-6

For those transitions should be considered.

Other packages found are listed here, although I assume they can be left
unchanged, or maybe just binNMUed:

  aria2 1.18.5-1
  bisonc++ 4.09.01-1
  cbmc 4.5-2
  clementine 1.2.0+dfsg-2
  fakeroot-ng 0.18-4
  fastx-toolkit 0.0.14-1
  chromium-browser -
  filezilla 3.8.0-1
  flexc++ 2.01.00-1
  grap 1.44-1
  imagevis3d 3.1.0-2
  jmtpfs 0.5-2
  kcov 11.1
  kdevelop-pg-qt 1.0.0-3
  mira 4.0.2-1
  mkvtoolnix 7.0.0-1
  natlog 1.01.0-2
  octave-image 2.2.1-1
  pingus 0.7.6-1.1
  python-casmoothing 0.1-2
  qtcreator 3.0.1-0
  ruby-passenger 4.0.37-3
  spring 96.0+dfsg-2
  sysdig 0.1.83-1
  tcpflow 1.4.4+repack1-2
  yrmcds 1.0.4-2


Proposing to make GCC 4.9 the default on the remaining non-release
architectures, then bumping the required g++ dependency in build-essential, then
file bug reports for the eleven packages mentioned in the first set, and then
transitioning these packages.

--- End Message ---
--- Begin Message --- 
On Fri, Oct 10, 2014 at 09:59:46 +0200, Emilio Pozuelo Monfort wrote:

> On 16/06/14 17:02, Matthias Klose wrote:
> >Package: release.debian.org
> >Severity: important
> >Tags: sid jessie
> >
> >This change mentioned in the GCC 4.9 news:
> >
> >   Improved support for C++11, including:
> >     The associative containers in <map> and <set> and the
> >     unordered associative containers in <unordered_map> and
> >     <unordered_set> meet the allocator-aware container
> >     requirements
> >
> >is an ABI incompatible change in the unordered_{map,set} classes built in c++11
> >mode.  The libstdc++6 library itself is not affected, so instead of having a
> >libstdc++6 transition, all packages providing a library built in c++11 mode and
> >using the unordered_{map,set} classes should be transitioned (changing the name
> >of the library package).
> >
> >A grep of all sources for #include *<unordered_{map,set}>, and then searching in
> >the log files for usage of -std=c++{0x,11} did show the following packages
> >building library packages (for packages not having verbose build logs I had to
> >look into the source package):
> >
> >   bobcat 3.22.00-1
> >   c++-annotations 9.9.1-2
> >   capnproto 0.4.0-1
> >   condor 8.0.6~dfsg.1-1
> >   liblo 0.28-3
> >   mrpt 1:1.0.2-1
> >   qtbase-opensource-src-gles 5.2.1+dfsg-1
> >   rapicorn 13.07.0~ds0-1
> >   shogun 3.2.0-2
> >   tesseract 3.03.03-1
> >   zeroc-ice 3.5.1-6
> >
> >For those transitions should be considered.
> 
> I'm not aware of any breakage because of this. Do you know if there's
> anything that actually broke because of that ABI break?
> 
Closing, as jessie is about to release.  Hopefully this won't bite us
later, but since nobody investigated the issue further there's not much
we can do at this point.

Cheers,
Julien

Attachment: signature.asc
Description: Digital signature


--- End Message ---
Reply to: