Joey Hess wrote: > And then, file is probably not the best test, the best test would be an > identical one to that used by dpkg-shlibdeps. There have been some bugs > because of this; not all of them are even fixed. #133092 is the best > one.
Very interesting. Hakan Ardo wrote: | When building a package that contains .so files intended for a non-native | architecture dh_shlibdeps will try to precess them and fail. Here is a | simple patch that will make dh_shlibdeps ignore such files: | | 72,73c72,73 | < $ff=file "$file"; | < if ($ff=~m/ELF/ && $ff!~/statically linked/) { | --- | > $ff=ldd "$file"; | > if ($ff !~ /not a dynamic executable/) { So is_elf() isn’t a good enough test for this use case. But the use case seems very iffy to me. I imagine that the tpkg-make tool allowed setting an arbitrary target for the cross-compilers it built. What about when the target happens to coincide with the host? Then you get dependencies you don’t want. Probably when the dust settles, dpkg-shlibdeps will learn to create appropriate multiarch dependencies, and paying attention to non-native binaries that have not been explicitly excluded will not seem to be a problem after all. Jonathan -- To UNSUBSCRIBE, email to debian-dpkg-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20100421163942.ga11...@progeny.tock