[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#325453: marked as done (developers-reference: Discourage use of -make clean in debian/rules)

Your message dated Fri, 29 Aug 2014 16:53:54 -0700
with message-id <20140829235354.GA23896@xanadu.blop.info>
and subject line Re: Bug#325372: lintian: please add check for careless usage of "make clean" and the like
has caused the Debian Bug report #325453,
regarding developers-reference: Discourage use of -make clean in debian/rules
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@bugs.debian.org
immediately.)


-- 
325453: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=325453
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems
--- Begin Message --- 
Package: lintian
Version: 1.23.10
Severity: wishlist

A lot of packages, particularly those with upstreams that use the GNU
autotools, have a very nasty habit of running the upstream clean rule while
utterly ignoring any errors from it.  This is very, very bad.  Please add a
check for it.

An egrep regex to catch this is probably about as simple as:

'\t[[:space:]]*-(\$[({]MAKE[)}]|make)[[:space:]].*(dist)?clean'

See transcript of IRC discussion below for further elaboration of rationale.

00:35 < Overfiend> joshk: What didn't you like about my changes to
   debian/rules?
00:36 < Overfiend> a configure target for autofoo-using packages is
   common practice AFAICT
00:36 < Overfiend> because of the problem of getting a properly cleaned
   tree
00:36 < Overfiend> if you were to assert that configure getting run
   *twice* for every package build from a freshly-unpacked source were
   a bit stupid, I'd agree.
00:36 < Overfiend> But that stupidity appears to be forced on us by
   GNU autofoo.
00:37 < joshk> yes - my scripts run a clean and then a build..
00:37 < Overfiend> your scripts?  debuild does that too.
   dpkg-buildpackage might as well, I don't remember.
00:37 < Overfiend> it's the Right Thing to do.
00:37 < vorlon> Overfiend: meh? Why would it be run twice?
00:37 < Overfiend> Builds should always start from a known state.
00:37 < Overfiend> vorlon: because debuild invokes debian/rules clean
   before debian/rules binary
00:38 < Overfiend> or dpkg-buildpackage does
00:38 < Overfiend> as I said, I don't remember
00:38 < doogie> why does clean depend on configure?
00:38 < vorlon> Overfiend: there's no reason clean should need to depend
   on configure
00:38 < Overfiend> can't "make clean" if there's no Makefile...
00:38 < vorlon> not with sane autotoolery
00:38 < doogie> then don't
00:38 < vorlon> yes, so you type "-make clean" instead
00:38 < Overfiend> UGH!
00:38 < doogie> what he said
00:38 < joshk> -$(MAKE) distclean, here
00:38  * Overfiend vomits across the room
00:38 < Overfiend> that's fucking moronic
00:38 < doogie> no, this is how it's done
00:38 < joshk> Overfiend: wow. have you considered entering a contest?
00:38 < Overfiend> "I don't care why make clean failed.  Ignore it."
00:39 < vorlon> fine.  [ ! -f Makefile ] || make clean
00:39 < vorlon> :P
00:39 < Overfiend> That's like swatting a fly with a neutron star.
00:39 < doogie> however, this is why I hate build systems that don't
   support a build dir.
00:39 < joshk> build system? what build system? :|
00:39 < Overfiend> vorlon: that's a bit more reasonable.
00:39 < joshk> vorlon: hmm, good idea
00:39 < doogie> ifeq (Makefile,$(wildcard Makefile))
00:40 < doogie>         make clean
00:40 < doogie> endif
00:40 < Overfiend> "Ignore all errors from make on the assumption that
   the only reason it would fail is because there's no Makefile" is
   mondo stupid.
00:40 < joshk> well, the reason i chose to do it my way is because it's
   that way in every single one of my autofoo using packages currently
00:40 < doogie> my way is more efficient.  less forking.
00:40 < joshk> so even if you were right, i'd end up having to orchestrate
   that change over all my packages anyway..
00:40 < joshk> IOW: someday.
00:41 < Overfiend> I don't care which approach, vorlon's or doogie's,
   people use.  but -$(MAKE) {dist,}clean is just unholy bad.
00:41 < vorlon> doogie: doesn't every line get passed to sh for execution?
   If [ is a builtin, there's no difference in fork count
   (Connection timed out)]
00:42 < Overfiend> Recall that people mindlessly ignoring error conditions
   is why we see most security advisories.
00:42 < doogie> vorlon: try again.  if the Makefile doesn't exist,
   $(wildcard) expands to nothing, so the line isn't sent to the shell
00:42 < vorlon> ah.
00:42 < vorlon> right.
00:42 < joshk> Overfiend: a ton of people, not just myself, use a
   variation of -make clean - you may want to submit a lintian check
   for it...
00:42 < doogie> and you have to test for the file's existance anyways.
00:42 < joshk> sounds like something that would get in. and it would
   compel me to do it more quickly :)
00:42 < doogie> altho, it does make the file harder to read
00:43 < Overfiend> joshk, doogie, vorlon: permission to quote you guys
   in my bug report.
00:43 < joshk> Overfiend: granted
00:43 < Overfiend> I'd rather paste this discussion than rewrite it.
00:43 < vorlon> Overfiend: as long as you leave out the fart noises
   [Read error: 104 (Connection reset by peer)]
00:43 < Overfiend> vorlon: oh, very well
00:43 < doogie> you can always quote my makefile foo, if it is meant to
   scare and intimidate
00:43 < Overfiend> doogie: :)

-- System Information:
Debian Release: testing/unstable
  APT prefers unstable
  APT policy: (500, 'unstable'), (500, 'stable')
Architecture: powerpc (ppc)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.9-powerpc-smp
Locale: LANG=C, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)

Versions of packages lintian depends on:
ii  binutils                   2.16.1-2      The GNU assembler, linker and bina
ii  diffstat                   1.41-1        produces graph of changes introduc
ii  file                       4.12-1        Determines file type using "magic"
ii  gettext                    0.14.5-2      GNU Internationalization utilities
ii  intltool-debian            0.30+20040213 Help i18n of RFC822 compliant conf
ii  man-db                     2.4.3-1       The on-line manual pager
ii  perl [libdigest-md5-perl]  5.8.7-4       Larry Wall's Practical Extraction 

lintian recommends no packages.

-- no debconf information

--- End Message ---
--- Begin Message --- 
Hi,

Using '-make clean' is now tested by lintian, and raises a warning:

W: viva source: debian-rules-ignores-make-clean-error line 32
N:
N:    A rule in the debian/rules file for this package calls the package's
N:    clean or distclean target with a line like:
N:
N:     -$(MAKE) distclean
N:    or
N:     $(MAKE) -i distclean
N:
N:    The leading "-" or the option -i tells make to ignore all errors.
N:    Normally this is done for packages using Autoconf since Makefile may not
N:    exist. However, this line ignores all other error messages, not just the
N:    missing Makefile error. It's better to use:
N:
N:     [ ! -f Makefile ] || $(MAKE) distclean
N:
N:    so that other error messages from the clean or distclean rule will still
N:    be caught (or just remove the "-" if the package uses a static
N:    makefile).
N:
N:    Severity: normal, Certainty: certain
N:
N:    Check: rules, Type: source

So I don't think that there's a need to mention this in dev-ref anymore, given
the widespread usage of lintian nowadays.

Thus closing this bug.

Lucas

--- End Message ---
Reply to: