Another step towards upstream, is to reuse the same gi.require_version
machinery.

In particular, create a helper which also provides a useful message if
the modules are out of date. Today this is only Debian/Ubuntu specific.

v2: Split from a larger commit (Emil)

All references bug one (in NEWS) was updated to the new domain, while
also bumping to httpS.

v2: Split from a larger commit, commit message, update all (Emil)

Add a couple of tweaks from upstream, getting us a step closer.

v2: Split from a larger commit, commit message (Emil)

From my (Emil) testing, there's no difference, yet I only use ASCII :-)
Regardless, we want to follow the style used by upstream.

v2: Split from a larger commit, commit message (Emil)

Seems like we're missing a couple of bits, while also diverging from
upstream.

v2: Split from a larger commit, commit message (Emil)
v3: Remove old __init__ (Andrej)

As of last commit we're in python2+3 land.

v2: Drop the TODO - tests are updated
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

These tests are not gtimelog specific, and out output format varies
across python versions.

If there was a short and trivial (read 3-5 lines of code) way to tweak
the test suite, we could have kept these. Yet cannot find out so, bye o/
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

The former is tad messy wrt python3, plus upstream has moved away from
using it.

v2: More meaningful commit message (Emil)

As spotted in our tests, the TZ entry is optional. Zero it to make
Python 3 happy.

v2: More meaningful commit message (Emil)

Either use the explicit "except ImportError", since that's what we care
about. Or drop the exception handling all together around socket, just
list upstream.

v2: More meaningful commit message (Emil)

With python3, the urllib module was split. Simultaneously we also need
to .encode() the output (of urlencode(), yeah I know) into bytes.

v2:
 - drop TEST tag, it's been tested quite a bit by now
 - squash extra urlencode hunk in rltimelog (thanks Ryan)
 - use explicit "except ImportError" (thanks Ryan)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Seemingly python2 is fine as-is, while python3 gets really unhappy

v2: Also handle the pickle.dump() codepath (thanks Ryan)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

The tzoffset code relies on python2 behaviour - aka using the / operator
will return a float. This is tad dubious, esp since // ensures that.

With python3 / returns a float, which we do not want here.

As we're here, make use of divmod() as per Marius's suggestion.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Wrap .items() and map() in a list() for iterating purposes. Even without
these the code works and produces the same output, modulo different
order in the parse_date_time dict() which is irrelevant.

Regardless, this was recommended by 2to3 so make it happy ;-)

v2: Also include categories.items()
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

With python3 dict.keys() is a "view" instead of a list. The shorthand
is sorted(dict).

At the same time, we use the removed cmp argument and cmp() built-in.

Switch to using sorted() + key=lambda
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

In these cases, there's little point in creating the object only to sort
it in-place immediately after. Instead use sorted() which creates a new
sorted object.

Flagged by 2to3, although I haven't checked if python3 complains as-is.

v2: Split categories sort() into different patch.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

From a quick observation, there is no notable functionality lost here.

Since sort() is no longer a thing (for dict) with python3, we opt for
sorted(). Yet that imperically doesn't quite work with python3.8. Quick
search suggests that with python3.7 dictionaries area already sorted.

As not much is lost here, while we do want python3 compat, drop the
sorting.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Former is removed with python3, while the latter is mostly the same.

In particular - this may cause minor increase in memory usage with
python2... Something we can live with at the moment.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Although in python2 this might be required, none of the actual text
actually needs it.

Could have left it as-is, although 2to3 was pretty insistent on removing
it.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Convert print to a function call print() to be compatible with python3.

Note: there are multiple others in the tests and misc scripts. We don't
care about those at the moment ;-)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Seems a far clearer, not to mention shorter to use the latter. As an
added bonus this addresses a handful of instances of unbracketed print

v2: Squash typo in _report() (thanks Ryan)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Otherwise with python3 they'll be a byte stream and we get into all
sorts of fun. This works like a charm across 2+3, albeit perhaps not
perfect on !linux platforms.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Quick look at the manual doesn't show e.streerror as a thing. Yet
python2.7 seems happy, while python3 isn't.

Rely on the implicit _str_ instead... if that one is busted we're out of
luck.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Namely s/,/as/, which seems also work with python2.6 or later... yay
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Some imports have been renamed with python3, while others moved through
the python3 cycle... Great job :-\

v2: use explicit "except ImportError" (thanks Ryan)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

This reverts commit 481182fe.

With all due respect, this is a semi-random combination of copy/paste.
As-is the branch is in foobar'd state and this patch doesn't help et al.

Converging with upstream get my +1000, although let's first have a
branch in usable state.

The follow-up commits do that, so let's revert this for now.

This reverts commit 27b38676.

Random cherry-picks are quite meh... Especially if they don't fix the
semi-broken state that the branch is in.

Namely s/,/as/, which seems also work with python2.6 or later... yay
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>

Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>