On 10 January 2017 at 04:52, Santiago Vila <[email protected]> wrote: > On Tue, Jan 10, 2017 at 04:20:10AM -0500, Jeremy Bicha wrote: >> On 10 January 2017 at 03:53, Santiago Vila <[email protected]> wrote: >> > Hi. Could you please write better changelog entries? >> > >> > Whoever reads that will think that the bug was about a new upstream >> > being uploaded, but that's definitely not the case. >> > >> > A better changelog would be for example like this: >> > >> > * New upstream release. >> > * Should fix FTBFS when using overlayfs (Closes: #842786) >> >> Actually, it doesn't fix anything because it's not clear that there's >> anything broken. > > To me it's quite clear what it's broken: It's the combination of a > source tarball not having directory entries, plus a kernel which does > not support overlayfs properly. > > We can fix the kernel, or we can fix the unorthodox tarball. > >> If the package weren't buildable, it should have >> shown up as unbuildable in the Reproducible Builds project. > > No, this is not black or white. They are using pbuilder, I'm using > sbuild. Packages buildable with pbuilder are not necessarily buildable > with sbuild, and viceversa. > > In fact, they do "dpkg-buildpackage", and I do "dpkg-buildpackage -A" > most of the time and sometimes "dpkg-buildpackage -B". Any of those is > a bug when it fails, but reproducible builds will not catch them if > they do just "dpkg-buildpackage". > > Also, they do not support shared memory, while I configured my sbuild > to support shared memory. Some packages I can build, they can't, > and you will see a strange error abour semaphores not working. > >> The >> changelog entry isn't completely useless as rebuilding the package (or >> uploading an updated package) proves that the package is buildable. > > Again, this is not black or white. > > And the changelog should not be used to close bugs if there is no > source changes associated with the closing. > >> Would it be more appropriate for me to just close this bug as invalid >> instead of doing it via changelog? > > If it was an invalid bug, yes, not only it would be more appropriate, > it would be certainly wrong to close it in the changelog. > (Just read the usual recommendations to write changelog entries). > > But this was not really the case. It was a perfectly valid bug. > > Why don't you just try to reproduce it? You just have to use sbuild > with overlayfs. > > Please try it, if you can't reproduce it I will try to help as much as > I can. > > Thanks.
I did try. As I reported on November 1, my build machine is Ubuntu, now updated to 17.04, using sbuild with union-type=overlay. I copied the get-orig-source rule from other Debian packges (see debian/rules for citations). If you think there's something here that could be done better, please submit a patch (and probably submit it to the other Debian packages too). Thanks, Jeremy
