Improve detection speed for LSB based systems

[roehling]

The new code avoids calling the external lsb_release
binary and reads the /etc/lsb-release issue file instead.

The old code called lsb_release quite excessively:

• once in is_os()
• twice in get_version() and get_codename()

On Ubuntu, this totals to 7 popen calls just to detect the OS.

[wjwwood]

Code and logic looks fine, but we should test it on lucid and quantal to ensure that nothing is different about the contents in /etc/lsb-release.

[roehling]

My work computer is quantal, and my file server at home is lucid. Both have the appropriate content in /etc/lsb-release.

[wjwwood]

Ok, sounds good to me then, @tfoote you want to take a look too?

[tfoote]

It looks like there's an lsb_release module so we don't have to do our own parsing:

>>> import lsb_release
>>> lsb_release.get_lsb_information()
{'RELEASE': '10.04', 'CODENAME': 'lucid', 'ID': 'Ubuntu', 'DESCRIPTION': 'Ubuntu 10.04.4 LTS'}

[roehling]

According to https://bugs.launchpad.net/rester/+bug/789736 we cannot assume that the lsb_release module is available on all systems. They suggest the platform module instead.

[tfoote]

I just ran into that too. It does appear the platform module does what we need and is more standard.

[roehling]

I tested the platform module code with Ubuntu and Debian, but I have no access to Linux Mint. As you can see, I had to change the Debian identifier to lower-case, so it might break on Linux Mint as well.

[jbohren]

Any chance a new version of rospkg with this fix could get released?

[tfoote]

This merge broke the unit tests. I think that #32 has everything required left. We're starting to transition toward requiring reviewed pull requests for commits to release branches. If you want to review #32 and +1 it for merging that would help.