control: retitle -1 dgit quilt-fixup incorrectly suggests using smash strategy
Hello, On Wed, Aug 17, 2016 at 05:46:49PM +0100, Ian Jackson wrote: > > For example: > > > > hephaestus ~/src/ocrmypdf % dgit quilt-fixup > > Format `3.0 (quilt)', checking/updating patch stack > > starting quiltify (multiple patches, linear mode) > > HEAD is now at ddd62dc release > > dgit: quilt fixup cannot be linear. Stopped at: > > dgit: 293e3b0e..4dabc96d: changed > > debian/patches/0001-patch-test-suite-executable.patch > > dgit: quilt fixup naive history linearisation failed. > > dgit: Use dpkg-source --commit by hand; or, --quilt=smash for one ugly > > patch > > If there are no changes then it shouldn't fail. But it may be that > there is some other problem and the bug is a wrong error message. > > Can you please send me all the ingredients ? I will need: > * All the .orig's involved > * The git HEAD you were working from Thank you for your reply. I now think that the bug is indeed the error message. When `dgit push` failed, I discovered that I had accidently committed a mode change to one of upstream's files. After I made a commit switching the mode back, dgit didn't complain that the fixup couldn't be linear, and correctly recognised that there was nothing quilty to commit. Since permission changes can't be represented by quilt patches, dgit could watch out for them and warn the user -- this would be a useful safety check enabled by dgit's HEAD==dsc requirement (perhaps suggesting using -D to see dpkg-source's output, as dgit push does). If you'd still like the ingredients to reproduce, you can `dgit clone ocrmypdf` and rewind head to e6eb64e, which is the commit before I undid the mode change. (You might have to wait for dinstall and/or check incoming.) -- Sean Whitton
signature.asc
Description: PGP signature
