Hi, Actually, I realize a more possible answer: do not depend on libstdc++6-X.X-dev.
I looked further, and libstdc++6-X.X-dev contain include files in /usr/include/c++/4.3/ which is used by the specific compilers only. The correct version of libstdc++6-X.X-dev is pulled in from g++-X.X, which is depended by g++, which is pulled in from build-essential; thus the correct version of libstdc++6-X.X-dev should be installed on all sane systems. At Wed, 5 Nov 2008 15:39:32 +0100 (CET), Andreas Tille wrote: > > On Wed, 5 Nov 2008, Junichi Uekawa wrote: > > > However, I'm not entirely happy with it, because you can satisfy > > build-dependency with any of the libstdc++6-X.X-dev, and does not > > provide a reasonable default, and it looks saner to use the latest > > version. > > That's what I would like to suggest: Always use the latest available > version. But "the latest" conflicts to a hard coded version in > d-shlibs. > > > Looking more closely, the development symlink (.so) is pointing to > > /usr/lib/libstdc++.so.6.0.10, but there is still opportunity to > > provide different include files; so that it can lead to subtle > > difference. I think the ABI is supposed to be the same since g++ 3.4, > > but that is modulo bugfixes, and I'm not sure if it's really sane to > > use libstdc++6-4.1-dev for the includes and link against the latest > > version of libstdc++ today. > > My reason for the hackish-workaround patch was that I just wanted > to clean up my system and remove everything conncected to "4.1" > versions og gcc. But I could not because my libraries ended up > depending from this old version while there was no reason not to > use the new version. IMHO even this alone is a good reason not > to hard code the version - if the ABI is not changed so we have > the good situation that packages do not need to be recompiled. > > > So, in summary; I don't quite know, I need more information. > > As I said: I'm defintely not able to provide more information - but > if I were you I would ask on debian-devel. If I'm not missleaded > I tried when I was adding the info to the bug report - but there > was no real answer if I remember right. So bringing it up again > is IMHO the best idea. > > Kind regards > > Andreas. > > -- > http://fam-tille.de > -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
