Bug#650445: [new check] "Multi-Arch: same" with files outside multiarch dirs

[Jakub Wilk <jwilk@debian.org>]

tags 650445 + patch
thanks

* Goswin von Brederlow <goswin-v-b@web.de>, 2010-01-29, 16:11:
> The Multi-Arch: same says the package can be installed for multiple 
> architectures at the same time. For that to work the package must have 
> no files with identical path/name for different architectures. The only 
> exception to this is when files are bit identical, which specifically 
> covers, but not limited to, the /usr/share/doc/package/* files required 
> in every package. Files that are bit identical between architectures 
> are usualy in /usr/share so that directory should be ignored completly 
> for this test. Files in $PATH may also be architecture independent if 
> they are scripts. Anything starting with "#!" should be ignored. I 
> think anything else that does not contain a triplet in its path/name 
> should be reported.

I propose the attached patch to detect arch-specific files outside 
multiarch-directories. It use quite different tactics than Goswin, but 
has an advantage of detecting most of the issues and producing no false 
positives (for the current state of the archive).

I ran lintian (with the patch applied) against all i386 multi-arch:same 
packages in the archive, and then compared it with what
http://people.debian.org/~jwilk/multi-arch/same-md5sums.txt reports. 
The only undetected problems were in the following files:
- shell scripts, *.desktop files, *.applications files and menu files with
embedded multi-arch paths[0];
- header files that differ across architectures;
- stuff in /usr/share/{doc,info,locale,lintian,man}/;
- some *.jar files that were generated at build time (and thus has 
timestamps embedded).

Of course, the patch description needs improvement. (Patches to the 
patch welcome! :P)

> Specificaly conffiles will pose a problem here I think. The conffile 
> will belong to multiple packages then causing the first problem for 
> dpkg. Then on conffile changes the change will happen multiple times, 
> once per architecture and the changes made by the first arch upgraded 
> will look like the user made changes to the conffile in other archs. 
> Overall a big mess.

Is it really how it's implement in dpkg? I seriously doubt it.

--
Jakub Wilk

diff --git a/checks/binaries b/checks/binaries
--- a/checks/binaries
+++ b/checks/binaries
@@ -128,7 +128,10 @@
 my $proc = shift;
 
 my $arch;
+my $multiarch;
 my $madir;
+my $gnu_triplet_re;
+my $ruby_triplet_re;
 my $dynsyms = 0;
 my $needs_libc = '';
 my $needs_libc_file;
@@ -140,6 +143,7 @@
 my %SONAME;
 
 $arch = $info->field('architecture')//'';
+$multiarch = $info->field('multi-arch')//'no';
 my $srcpkg = $proc->pkg_src()//'';
 
 foreach my $file (sort keys %{$info->objdump_info}) {
@@ -184,6 +188,12 @@
 # nsswitch modules
 $madir = $MULTIARCH_DIRS->value($arch);
 
+$gnu_triplet_re = quotemeta $madir;
+$gnu_triplet_re =~ s,^i386,i[3-6]86,;
+$ruby_triplet_re = $gnu_triplet_re;
+$ruby_triplet_re =~ s,linux\\-gnu$,linux,;
+$ruby_triplet_re =~ s,linux\\-gnu,linux\\-,;
+
 # In the (unlikely) case that the architecture is unknown (or we do
 # not know the multi-arch path for a known architecture) , we assume
 # it the multi-arch path to be this (hopefully!) non-existent path to
@@ -267,6 +277,12 @@
         tag 'arch-dependent-file-in-usr-share', $file;
     }
 
+    if ($multiarch eq 'same') {
+        unless ($file =~ m,/(?:$gnu_triplet_re|$ruby_triplet_re|java-\d+-openjdk-\Q$arch\E)/,) {
+            tag 'arch-dependent-file-not-in-arch-specific-directory', $file;
+        }
+    }
+
     # ELF?
     next unless $fileinfo =~ m/^[^,]*\bELF\b/o;
 
diff --git a/checks/binaries.desc b/checks/binaries.desc
--- a/checks/binaries.desc
+++ b/checks/binaries.desc
@@ -113,6 +113,13 @@
 Info: This package installs an ELF binary in the <tt>/usr/share</tt>
  hierarchy, which is reserved for architecture-independent files.
 
+Tag: arch-dependent-file-not-in-arch-specific-directory
+Severity: serious
+Certainty: possible
+Ref: https://wiki.ubuntu.com/MultiarchSpec
+Info: This package is Multi-Arch "same", but it installs an ELF binary in the
+ directory that is not architecture-specific.
+
 Tag: binary-in-etc
 Severity: serious
 Certainty: certain