On 04-Oct-25 11:16, Daniel Jacobowitz wrote: > On Mon, Oct 25, 2004 at 05:06:21PM +0200, Andreas Jochens wrote: > I really don't agree that doing this based on the performance issues > makes sense. And you're wrong about mips64; the most efficient > libraries on mips64 targets are usually n32, which live in /lib32 > rather than the o32 binaries in /lib. I don't know about s390x. > Or about ppc64, which does the same.
The Debian amd64 port installs 64bit libraries in '/lib'. This is a fact and this will not change. The _only_ place where the '/lib64' directory which is a symlink to '/lib' is used by the Debian amd64 port is the name of the interpreter which is specified in gcc. This makes the whole system depend on the '/lib64' symlink an I want to avoid that. Users have accidently removed that symlink (sometimes by trying to install external software) which made the system completely unusable. I want to get rid of the dependency on the '/lib64' symlink to avoid that effect and to make further changes to run external software easier (e.g. by making /lib64 a separate directory - such a change by a package upgrade will be nearly impossible as long as 'ln' depends on that symlink). > distributions. Of course recompiling the whole distribution works; > it's only a compatibility issue. So I don't understand what that is > supposed to show. It is supposed to show that the interpreter name is the only place where the '/lib64' symlink is really used. This is not trivial, because in theory lots of packages might already check for the '/lib64' during the build process and could either FTBFS or stop running if the '/lib64' (or '/usr/lib64') directory is not present. However, this is not the case. It is supposed to show that we can have the well known standard unix filesystem layout on amd64 without any '/lib64' and symlink tricks. > > Do you really count this patch as a "radical" departure from a > > standard? > > Yes. I do not think the commercial distributions will provide a > symlink for compatibility with Debian, unless there is official ABI > documentation showing that this is the way to go. That's where you > should start. Come on, be realistic. If I cannot convince Debian to start using an interpreter name in '/lib', it will be diffcult to convince somebody from the commercial distributions to install the symlink. It is a chicken and egg problem and you know that. However, at least Gentoo and Ubuntu (and maybe others) already have the necessary symlink. > Please, tell me - has anyone discussed this compatibility issue outside > of Debian? There have mainly been talks about general multiarch support which have in common that the interpreter is located in '/lib': http://www.linuxbase.org/~taggart/multiarch.html specifies /lib/x86_64-linux/ld.so.2 as the interpreter name for multiarch. (Would you accept that one?) If my arguments do not convince you that to get rid of the '/lib64' dependency now, i.e. before Debian starts to officially support amd64, would be a good move, please feel free to close this bug report. Regards Andreas Jochens
