Hi, On Fri, 28 Dec 2018 09:18:36 +0100 Johannes Schauer <jo...@debian.org> wrote: > On Sun, 23 Dec 2018 12:56:55 -0500 Michael Stone <mst...@mathom.us> wrote: > > Please just wait > I take that means that you intend to apply the fix I proposed? > > The problem now became more dire for fakechroot because mv(1) is now used in > the maintainer script of dash (during the diversion process). So as of a few > days ago it is now impossible to use debootstrap with fakechroot.
I'm sorry, but my attached patch was apparently not properly tested by me. :( More work is needed to backport the patch from upstream to version 8.30 because just calling renameat2 as proposed in the patch I attached in my first message will just result in an infinite recursion loop. I guess I only didn't notice because the initial patch never worked because HAVE_RENAMEAT2 never got defined as the following hunk was missing from it: --- coreutils-8.30.orig/lib/config.hin +++ coreutils-8.30/lib/config.hin @@ -2069,6 +2069,9 @@ /* Define to 1 if you have the `renameat' function. */ #undef HAVE_RENAMEAT +/* Define to 1 if you have the `renameat2' function. */ +#undef HAVE_RENAMEAT2 + /* Define to 1 if you have the `rewinddir' function. */ #undef HAVE_REWINDDIR I only now started to understand how the Debian package includes the gnulib submodule and ends up shipping a number of autogenerated files as well. Upstream doesn't have the problem of infinite recursion because they renamed renameat2 to renameatu. So one way of solving this would be to also backport that change. Would you be okay with that or would you like to see another solution? How much time do I have to present a working patch? Thanks! cheers, josch
signature.asc
Description: signature
