Author: Richard
[Added to the original trac issue at 12.59pm, Monday, 4th July 2011]

There's a zero-length way in the XML, which shouldn't have got in there. Any ideas how that might have happened? (it's usually something to do with undo...)

Author: DaveF
[Added to the original trac issue at 2.02pm, Monday, 4th July 2011]

Replying to [comment:1 Richard]:

There's a zero-length way in the XML, which shouldn't have got in there. Any ideas how that might have happened? (it's usually something to do with undo...)

Which nodes are those? I can't see any that are the same. I didn't start undoing until I got the error. I was using the parallel command on closed ways & placing them on top of each other to create near identical duplicates. Isn't there a way to ignore 0 lengths? Didn't P1 do that? There is a bot that cleans up duplicate points so they must be getting uploaded somehow.

Author: Richard
[Added to the original trac issue at 2.12pm, Monday, 4th July 2011]

The line

<way id="-37" version="0" changeset="8628320"/>

is a way without any nodes in it.

In theory P2 could do a last-minute check before uploading, and remove them from the changeset before sending (that's roughly what P1 did), but that's not really fixing the problem - it shouldn't be creating 0-length ways internally in the first place. I'd like to find out what circumstances make it do so.

Author: DaveF
[Added to the original trac issue at 2.39pm, Monday, 4th July 2011]

Replying to [comment:3 Richard]:

The line

<way id="-37" version="0" changeset="8628320"/>

is a way without any nodes in it.

Oh, I see. That's not actually a way with zero length (ie two nodes with identical lat & lon,) but poorly constructed xml.

If P2 would of excepted it without that line then it must be better/easier (especially for the end user) if it did some verification. Leaving major errors like that will only discourage users from adding to OSM

Author: DaveF
[Added to the original trac issue at 5.27pm, Monday, 4th July 2011]

Replying to [comment:4 DaveF]:

If P2 would of excepted it without that line...

Oh dear; apologies for the poor spelling, That, of course, should have been 'accepted'.

Author: Ollie
[Added to the original trac issue at 4.40pm, Monday, 8th August 2011]

I think I've found a reproduceable method of getting nodeless ways, so leading to this error.

1. Create a way.
2. Press "P" to start creating a parallel way.
3. Without moving the mouse (i.e. before the new-extent lines start to appear), click.

I often do something similar to this (although I move the mouse back and forth to insubstantiate the way properly) in order to create a "duplicate" identically-shaped object, e.g. blocks of houses. If I don't move the mouse, the problem occurs.

Below is the OSM.xml which got created when I did the above, and so causes the error to appear.

Author: Richard
[Added to the original trac issue at 1.37pm, Monday, 3rd October 2011]

Ollie - thanks; have fixed that in systemed/potlatch2@dd2f112 . Am closing this for now, but if anyone can find another way of creating 0-length ways then they can reopen it.