revtreeview:load_visible_revisions() is treating revids as a unicode string

[John Szakmeister]

I found this by running qlog against a Subversion repo. It turns out that load_visible_revisions() is yanking the revid out of a PyQt construct and storing it in a set(). The revision ids eventually make their way down to svn/repository.py:get_revision() where it asserts that the revid is a string. At that point, I get an error box saying it was an invalid revision id. We should be treating that string as bytes, rather than a unicode string. I did not check to make sure that it gets into the index correctly (as bytes).

[John Szakmeister (jszakmeister)]

This looks like it's only a problem in the 0.14.x line, as trunk has this:
            if not revid.isNull():
                revids.add(str(revid.toByteArray()))

[John Szakmeister (jszakmeister)]

Yep, looks like Lukas ran into the issue. It was fixed in r933 of trunk. Perhaps that could be ported to the 0.14.x line?

[Alexander Belchenko (bialix)]

John, the backport patch will be greatly appreciated.

[John Szakmeister (jszakmeister)]

On Fri, Oct 16, 2009 at 3:08 PM, Alexander Belchenko <email address hidden> wrote:
> John, the backport patch will be greatly appreciated.

I'll take care of that later tonight. Thanks Alexander!

-John

[John Szakmeister (jszakmeister)]

Here's a patch with the backport of Lukas's revision id fix to the 0.14 branch.

[Alexander Belchenko (bialix)]

Thanks!