[Patch] NullPointerException in WayConnectedToArea due to Way w/ 0 nodes

[anonymous]

Repository Root: http://josm.openstreetmap.de/svn
Build-Date: 2012-01-30 02:32:23
Last Changed Author: stoecker
Revision: 4878
Repository UUID: 0c6e7542-c601-0410-84e7-c038aed88b3b
URL: http://josm.openstreetmap.de/svn/trunk
Last Changed Date: 2012-01-30 00:02:54 +0100 (Mon, 30 Jan 2012)
Last Changed Rev: 4878

Identification: JOSM/1.5 (4878 de)
Memory Usage: 137 MB / 494 MB (31 MB allocated, but free)
Java version: 1.7.0, Oracle Corporation, Java HotSpot(TM) Client VM
Operating system: Windows 7

Dataset consistency test:
[WARN - ZERO NODES] Way {Way id=152739257 version=2 MVT nodes=[]} has zero nodes
[WARN - ZERO NODES] Way {Way id=110239757 version=10 MVT nodes=[]} has zero nodes



java.lang.NullPointerException
	at org.openstreetmap.josm.data.validation.tests.WayConnectedToArea.visit(WayConnectedToArea.java:27)
	at org.openstreetmap.josm.data.osm.Way.visit(Way.java:187)
	at org.openstreetmap.josm.data.validation.Test.visit(Test.java:139)
	at org.openstreetmap.josm.actions.upload.ValidateUploadHook.checkUpload(ValidateUploadHook.java:60)
	at org.openstreetmap.josm.actions.UploadAction.checkPreUploadConditions(UploadAction.java:137)
	at org.openstreetmap.josm.actions.UploadAction.checkPreUploadConditions(UploadAction.java:103)
	at org.openstreetmap.josm.gui.io.SaveLayersDialog$SaveAndUploadTask.uploadLayers(SaveLayersDialog.java:417)
	at org.openstreetmap.josm.gui.io.SaveLayersDialog$SaveAndUploadTask.run(SaveLayersDialog.java:521)
	at java.lang.Thread.run(Unknown Source)

[simon04]

IMHO it makes sense to skip all validation tests for ways with 0 or 1 nodes as those objects shouldn't be there anyway and presumably also other tests may be broken (cf. #7437). See attached patch.

What do you think?

[akks]

I tried to make the same for #7437 in [5012], but does not know if it was helpful. But there was no bad consequences :)

[skyper]

Replying to akks:

I tried to make the same for #7437 in [5012], but does not know if it was helpful. But there was no bad consequences :)

I think we should rather find the reason for these zero node ways than implementing work arounds for every occasion where these ways cause problems !

1. We definitly have a problem with objects mentioned twice in the validator list.
2. As the later problems with these zero node ways seem to have a different source. As I have seen a similar strange exception last week merging layers I think the problem lies somewhere around downloading/merging/conflicts solution.

Sorry, had not time the other day to check and it was not reproducable right away.

[simon04]

Replying to akks:

I tried to make the same for #7437 in [5012], but does not know if it was helpful. But there was no bad consequences :)

This patch tries to take a more general approach as I assume that many validation tests rely on ways having at least 2 nodes (i.e., getting consistent data). IMO it should not be the task of each individual validation test to check that the data is consistent.

[simon04]

#5764 is (somehow) related.

[skyper]

Ticket #7763 has been marked as a duplicate of this ticket.

[simon04]

I agree that the source of these inconsistent objects should be spotted eventually. However, for a less buggy JOSM, I suggest to apply the above patch …

[mrwojo]

Re the patch: I suggest moving the getNodesCount() check to a separate method that individual tests can override. UntaggedWay wants these 0-1 node ways.

[mrwojo]

Ticket #7912 has been marked as a duplicate of this ticket.

[bastiK]

Replying to mrwojo:

Re the patch: I suggest moving the getNodesCount() check to a separate method that individual tests can override. UntaggedWay wants these 0-1 node ways.

Sounds reasonable.

[stoecker]

Status of this?

[Don-vip]

Ticket #8167 has been marked as a duplicate of this ticket.

[Don-vip]

There's another test file in #8167, if needed.

[Don-vip]

Ticket #7562 has been marked as a duplicate of this ticket.

[Don-vip]

Ticket #8462 has been marked as a duplicate of this ticket.

[Don-vip]

Ticket #8569 has been marked as a duplicate of this ticket.

[Don-vip]

There's another test file in #8569, if needed.

[rickmastfan67]

Just a thought here, but #7986 and #8137 give a full-proof way to create 0-node ways. Sure, the crashes are in a slightly different place (2 lines off), but it's crashing in the same exact files. If they could be fixed, it might help fix this bug as well.

[Don-vip]

Thanks, I have started to look into undelete plugin code, see ​[o29474].

[Don-vip]

Ticket #8630 has been marked as a duplicate of this ticket.