drop utf-8 encoding to compute file hashes in symlink stage

[rhaschke]

This fixes #311 (UnicodeDecodeErrors) as pointed out in #311 (comment). Relates to #368.

[wjwwood]

@rhaschke there are CI failures for this. I've merged #368, but I don't have push permission on this branch so I cannot rebase it (I'm assuming rebasing would address the issue). Can you rebase this or otherwise fix these failures (assuming they're related to the changes). Thanks.

[rhaschke]

@wjwwood: I rebased. However, Travis already fails on master. I fixed this in #428.

[rhaschke]

@wjwwood: I finally digged into this and found the underlying issue:
@jbohren's idea was to compute the md5 hashes of files to compare them. However, opening files, opens them in text-mode by default, returning string objects when calling read(), which is not compatible to digest() that expects a byte array. Recoding sometimes failed on binary files. My first fix only removed the recoding to utf8. However it's also import to read the file in binary mode in the first place.
Travis passes now. This resolves #433 too (which is a rebase of #399).

[mikepurvis]

Nice!

[wjwwood]

Awesome, thanks @rhaschke.

[furushchev]

Hi, is there any plan to release new version including this patch?

[wjwwood]

Hi, is there any plan to release new version including this patch?

Yes, I'm just swamped. But I'm going over the open pr's and issues now, so I think a new release is incoming. Sorry for the long drought of updates.

[nckswt]

Any update on the release of 0.4.5? Getting this fix would help out a lot!

[HareshKarnan]

sudo apt-get install python-catkin-tools fixed things for me !