[mkgmap-dev] Occasional header corruption in gmapsupp.img

Sat Nov 29 07:42:06 GMT 2014

Hi all,

once again I stumbled over a really weird problem:

In rare cases mkgmap produces a gmasupp.img that shows up as "Map data (c)
OpenSt" in the list of maps on my Oregon 550.
Screenshots:
<http://gis.19327.n5.nabble.com/file/n5825759/bad_01.png> 

<http://gis.19327.n5.nabble.com/file/n5825759/bad03.png> 

<http://gis.19327.n5.nabble.com/file/n5825759/bad04.png> 

<http://gis.19327.n5.nabble.com/file/n5825759/good_01.png> 

<http://gis.19327.n5.nabble.com/file/n5825759/good03.png> 

<http://gis.19327.n5.nabble.com/file/n5825759/good04.png> 

There is no error message or warning when building the gmapsupp.img,

I observe this problem for some 4 month now and I can 100% reproduce the
bug.

It seems to be triggered by the OSM-inputdata although the OSM-data itself
is OK.

The problem occurred on a DACH-extract with my bikeroute-highlights-overlay
style a few months ago,
now I have it with a ITALY-extract and my housenumbers-overlay style.

Usually I build 6 different gamapsupp.img files for 6 different styles from
the same tiles.
Only one of them, if any, has the buggy header.

The problem usually disappears after an update of the OSM-data, just to
reappear a few days or weeks later.

I kept 2 versions of the ITALY-extract for reference and with those I can
100% reproduce a faulty gmapsupp.img.

I often update the versions of splitter and mkgmap I use for building the
maps.
mkgmap-r3356 and splitter-r414 still produce the same bug.

The problem assembling a test-case is that the ITALY-extract has a size of
1.2 GB .
When I remove just one tile from the split extract the gmapsupp.img is OK.
I cannot produce a small testcase.

I uploaded a sample of a gmapsupp.img with the bug, and one that is error
free.
http://files.mkgmap.org.uk/download/233/maps.tgz

Ciao,
  Franco

--
View this message in context: http://gis.19327.n5.nabble.com/Occasional-header-corruption-in-gmapsupp-img-tp5825759.html
Sent from the Mkgmap Development mailing list archive at Nabble.com.