[OSM-dev] Possible Potlatch Bugs (was: Daily Diff Merging)

D Tucny d at tucny.com
Thu Oct 25 03:56:15 BST 2007


On 25/10/2007, Richard Fairhurst <richard at systemed.net> wrote:
>
> Brett Henderson wrote:
>
> > The history is a bit odd. The sequence of events appears to be:
> > * Node is created.
> > * Node is deleted immediately followed by way creation referring to
> > it.
> > * Way and node are both updated but node is still deleted (ie. not
> > visible).
> >
> > Can anybody shed some light on why this is occurring?
>
> I'm continuing to look into this one. It may well be connected with
> the repeatedly mentioned bug whereby a "whichways" call (i.e. get
> POIs) can take place halfway through a "putway" call (i.e. save way)
> - in other words, when the nodes have been written but not the way,
> causing whichways to identify them as POIs. The user then sees these
> misleading POI nodes and tries to delete them. I have tried to squash
> this bug but haven't managed yet.
>
> Anyone else is very welcome to look over the Potlatch source, it
> doesn't just have to be me... :)


This explanation sounds plausible,  the mapper in question here is a new
mapper, mostly using potlatch, one thing to note too, his latency will be
massive compared to your test environment with a splashing of packet loss to
help, probably along the lines of 300ms latency plus average 10% packet
loss... As a result, actions performed will likely take a lot longer to
commit than you will typically see and some of them will fail with
connection errors, so will need retrying, all adding delays into the process
though... If you have a look at the data in the area in question now, you
can see some of the issues this user is having... I've been giving him some
pointers and I'm going to go through the data in the next few days to get
some more information to him regarding specific things that have been
created, but, I don't think it's all things he's knowingly done... e.g.
overlapping ways with different tags, ways seemingly being extended from
part way through an existing way creating a new way...
He seems to be putting a lot of effort into this, which can't be knocked,
but, his work is currently marred by all sorts of bad data that he's
probably not knowingly entering...

If you look at the area, not the river running through, this was pulled in
from the pgs data and tidied up by myself, it looks like he's tried to
improve it in some places, the result being that some of the coastline ways
is now pointing in the wrong direction, some ways have been duplicated and
there's a heavy spattering of unused nodes around these ways... Looking
wider, there's large amounts of overlapping data, nodes, ways etc... It's
quite a mess, I'll be working on cleaning it up as best I can and will be
talking the the user that created it to see if we can find a way to prevent
these things happening more, this is going to take quite a lot of time, but,
I want to help him get things right and don't want him to just give up
because the mess is unmanagable... As I said before, I'd guess most of this
is due to delays, either resulting in the user redoing things that don't
seem to have taken effect, or, potlatch doing similar...

If you want someone to do some tests over crappy international internet
connectivity, let me know... I'm no flash developer, but, if you want, I can
run through some tests for you and feedback the results to you...

d
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20071025/b706fba1/attachment.html>


More information about the dev mailing list