On Sat, Jan 07, 2017 at 02:41:40PM +0100, Johannes Schauer wrote: > Quoting Santiago Vila (2017-01-07 13:37:27) > > On Sat, Jan 07, 2017 at 01:02:19PM +0100, Matthias Klose wrote: > > > You could detect these as well by inspecting the Built-Using flag > > > and using the very same versions for your test builds. > > > > Well, maybe this is a bug in sbuild. > > > > I'm Cc:ing Johannes Schauer, sbuild maintainer, in case he has any > > idea about how to avoid this minor glitch. > > even after having read the bug I'm unsure what the problem with sbuild is > here. > > What do you want to use the Built-Using header for? What is sbuild not doing > that you think it should do?
This package, cross-toolchain-base, has a small collection of Build-Using fields on other source packages. I'm building packages in stretch, and when a package FTBFS in stretch, I report it as a serious bug. Matthias complains that I'm not building packages "the same way" he did. He seems to suggest (please correct me if I'm wrong) that I should use the exact same version he used in the Built-Using field. But I think this information is only in the binary packages. So: Is sbuild supposed to take this information from the binary packages I'm trying to rebuild (unlikely, but I ask just in case), or maybe it's that the Built-Using fields should just be versioned when there is a risk that not every version of the Built-Used packages is ok for the build? (May Built-Using be versioned, like Depends or Build-Depends?) [ My theory is that this is a bug in either cross-toolchain-base, sbuild, or maybe our package format, but definitely not in myself building packages on stretch ]. Thanks.
