Matching package names across distributions

[Enrico Zini]

On Wed, Feb 02, 2011 at 05:36:18PM -0500, James Antill wrote:

> % ./distromatch/distromatch fedora xapian-bindings-python     
> fedora:xapian-bindings-python debian:
> fedora:xapian-bindings-python mandriva:xapian-bindings-python
> fedora:xapian-bindings-python suse:

Fixed, by adding regexp to consider installed python modules as
"interesting" files (thus relevant for matching packages):

  $ ./distromatch fedora xapian-bindings-python
  fedora:xapian-bindings-python debian:python-xapian
  fedora:xapian-bindings-python mandriva:xapian-bindings-python
  fedora:xapian-bindings-python suse:

> % ./distromatch/distromatch fedora glibc
> fedora:glibc debian:
> fedora:glibc mandriva:glibc
> fedora:glibc suse:glibc

Here I fixed the shlib regexp to include libraries in /lib (previously
it only considered libraries in /usr/lib):

  $ ./distromatch fedora glibc
  fedora:glibc debian:hurd,ia32-libs-core,libc0.1,libc0.1-i386,libc0.3,libc6,libc6-amd64,libc6-i386,libc6-mips64,libc6-mipsn32,libc6-ppc64,libc6-s390x,libc6-sparc64,libc6.1
  fedora:glibc mandriva:clauer,glibc,ld.so1
  fedora:glibc suse:glibc

I hadn't realised Debian has so many glibc packages :)

Mandriva shows a couple of false positives I'll need to investigate.

Implementing matching of source packages could probably help in this
case, by restricting the output to those binary packages generated by
matching source packages. Matching source packages may require more
input data sources, though.

> % ./distromatch/distromatch fedora zsh-html
> fedora:zsh-html debian:
> fedora:zsh-html mandriva:
> fedora:zsh-html suse:

Is zsh-html the documentation of zsh, therefore something like
http://packages.debian.org/sid/zsh-doc ?

> % ./distromatch/distromatch fedora openoffice.org-calc
> fedora:openoffice.org-calc debian:app-install-data,broffice.org,libreoffice-calc,openoffice.org-calc
> fedora:openoffice.org-calc mandriva:openoffice.org-calc
> fedora:openoffice.org-calc suse:

This is hit by the openoffice->libreoffice transition. I'll need to
investigate a bit more in the next days.

> % ./distromatch/distromatch fedora xpaint
> fedora:xpaint debian:app-install-data,kappfinder,xpaint
> fedora:xpaint mandriva:kappfinder,libxaw3dxft6,xpaint
> fedora:xpaint suse:

Restricting matching .desktop files to only those in
/usr/share/applications gets rid of some false positives. Not sure about
libxaw3dxft6, will have to investigate as well.

  $ ./distromatch  fedora xpaint
  fedora:xpaint debian:libxaw3dxft6,xpaint
  fedora:xpaint mandriva:lib64xaw3dxft6,libxaw3dxft6,xpaint
  fedora:xpaint suse:

> % ./distromatch/distromatch fedora openssl
> fedora:openssl debian:openssl
> fedora:openssl mandriva:libopenssl0.9.8,libopenssl1.0.0,openssl
> fedora:openssl suse:compat-openssl097g,libopenssl0_9_8,openssl

This now has even more false positives after I've improved library
matching:

  $ ./distromatch  fedora openssl
  fedora:openssl debian:ia32-libs,libssl0.9.8,libssl1.0.0,openssl
  fedora:openssl mandriva:games-compat,lib64openssl0.9.8,lib64openssl1.0.0,libopenssl0.9.8,libopenssl1.0.0,openssl
  fedora:openssl suse:compat-openssl097g,libopenssl0_9_8,libopenssl1_0_0,openssl,openssl-doc

I need to add more filtering of the results of fuzzy match algorithms.

> ...it also doesn't seem to understand multilib. so you can find out that
> blah.x86_64 on Fedora is libblah62.x86_64 on SuSE, but you can't find
> out that blah.i686 is libblah62-32bit.x86_64.

Can you give me an example of a specific value of blah?

Anyway, you can pull again from git to get the results of my work of
today.


Ciao,

Enrico

-- 
GPG key: 4096R/E7AD5568 2009-05-08 Enrico Zini <enrico at enricozini.org>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.freedesktop.org/archives/distributions/attachments/20110205/4e4bbaea/attachment.pgp>