On 2016-10-23 11:51-0700, Sean Whitton wrote: > You should use "Forwarded: not-needed" (see DEP-3). This does not seem to work with gbp-pq (see #785274), I propose to add this as soon as gbp-pq supports DEP-3. >>> 2. You can fix all of these Lintian tags, except possibly >>> hardening-no-fortify-functions. You should definitely deal with the >>> warnings. >>> >>> W: patat-dbgsym: debug-file-with-no-debug-symbols >> >> I've updated debian/rules to something matching >> stylish-haskell. > > Okay. I'll take a proper look soon. Thanks! >>> I: patat: spelling-error-in-binary usr/bin/patat Nam Name >>> I: patat: spelling-error-in-binary usr/bin/patat isn't isn't >>> I: patat: spelling-error-in-binary usr/bin/patat forward forward >>> I: patat: spelling-error-in-binary usr/bin/patat upto up to >>> I: patat: spelling-error-in-binary usr/bin/patat discontigous discontiguous >>> I: patat: spelling-error-in-binary usr/bin/patat uncomplete incomplete >>> I: patat: spelling-error-in-binary usr/bin/patat The The >> >> Not sure about this one... Is "patat" too generic for lintian? I've >> added this to debian/lintian-overrides. > > I don't understand. It is pointing out misspellings, such as > 'uncomplete', somewhere in the upstream source. You can add a quilt > patch to fix them, and forward it upstream. As I didn't found anything matching these errors in the source, I thought it was a generic error message concerning the binary name. Now, that I understood the purpose of this check, I can only found these mistakes in the binary itself, so I guess these are in the dependencies... >>> I: patat: hardening-no-bindnow usr/bin/patat I: patat: >>> hardening-no-pie usr/bin/patat >>> >>> I think that in order to pass hardening options to gcc, if you're >>> willing to work on that, you'll need to abandon the CDBS build system >>> you're using at present. See the Makefile for keysafe[1] (not yet in >>> Debian) to see how to pass the options, and the rules file for the >>> stylish-haskell package to see how to do without CDBS. >> >> After reading this Makefile, I'm not sure how keysafe avoids >> hardening-no-bindnow and hardening-no-pie... Do you have any clue? > > The Makefile propagates LDFLAGS, CFLAGS and CPPFLAGS through to ghc. > Then you enable all hardening in your d/rules,[1] and the right flags > get set by debhelper. > > [1] https://wiki.debian.org/Hardening I would like to wait a little before adding this: the default flags added to gcc seems quite new, so I propose to have a look again when things stabilize. >>> 3. Please run upstream's test suite during the package build. >> >> Should be done now, I'm not sure about how I run tests... See >> debian/rules override_dh_auto_test > > Okay, I'll look later. Thanks. >> Concerning the last lintian warning (binary-without-manpage), I'm not >> sure it will add anything to "--help", and it is going to add an >> overhead to maintain the package... If it's not required I would >> prefer not to do anything with this. > > Adding manpages is one of the things that Debian does to standardise and > bring together the different programs that are part of the Debian > system. I'd strongly encourage you to be part of this QA effort. > > With regard to maintenance, hopefully you can persuade upstream to adopt > your manpage, so there wouldn't be any overhead. > > In the meantime, you can use help2man to generate the manpage. Note > that you shouldn't run help2man during the package build, as it's a > notorious source of FTBFS bugs. Instead, add a target to d/rules to > generate the manpage. See the ocrmypdf source package's d/rules. > > If help2man is insufficient, see again stylish-haskell where I use > asciidoctor. I'll try to add a manpage using help2man. Concerning DHG's package-plan, I can't run the tests myself, ghc seems to be broken in my chroot due to hardening flag -pie (see #712228). So I propose to add patat later, when things calm down.
Attachment:
signature.asc
Description: PGP signature