Passing GeoJSON directly to ogrinfo causes a segfault

[homme]

The page at ​http://www.gdal.org/ogr/drv_geojson.html states that a GeoJSON datasource can be "Text passed directly and encoded in GeoJSON". Trying this with the following two ogrinfo commands causes a segfault:

ogrinfo '{"type":"Feature","properties":{"foo":"bar"},"bbox":[-180.0,-90.0,180.0,90.0],"geometry":{"type":"Polygon","coordinates":[[[-180.0,10.0],[20.0,90.0],[180.0,-5.0],[-30.0,-90.0]]]}}'
ERROR 4: GeoJSON Driver doesn't support update.
Had to open data source read-only.
INFO: Open of `{"type":"Feature","properties":{"foo":"bar"},"bbox":[-180.0,-90.0,180.0,90.0],"geometry":{"type":"Polygon","coordinates":[[[-180.0,10.0],[20.0,90.0],[180.0,-5.0],[-30.0,-90.0]]]}}'
      using driver `GeoJSON' successful.
Segmentation fault

ogrinfo '{ "type": "Point", "coordinates": [100.0, 0.0] }'
ERROR 4: GeoJSON Driver doesn't support update.
Had to open data source read-only.
INFO: Open of `{ "type": "Point", "coordinates": [100.0, 0.0] }'
      using driver `GeoJSON' successful.
Segmentation fault

I am using GDAL 1.7.0, released 2010/01/19 on a Gentoo system (Linux peaches 2.6.31-gentoo-r6 #2 PREEMPT Tue Dec 22 13:26:44 GMT 2009 i686 AMD Athlon(tm) XP 2800+ AuthenticAMD GNU/Linux) compiled using gcc (Gentoo 4.3.4 p1.0, pie-10.1.5) 4.3.4.

I mention the two commands because they seem to produce different backtraces in gdb. The backtrace for the first is:

#0  0x4f2ccdd3 in *__GI___strcasecmp (s1=0x8051428 "{ \"type\": \"Point\", \"coordinates\": [100.0, 0.0] }", s2=0x0) at strcasecmp.c:65
        loc = <value optimized out>
        result = <value optimized out>
#1  0x0804a023 in main (nArgc=2, papszArgv=Cannot access memory at address 0x4
) at ogrinfo.cpp:214
        pszWHERE = 0x0
        pszDataSource = 0x8051428 "{ \"type\": \"Point\", \"coordinates\": [100.0, 0.0] }"
        papszLayers = 0x0
        poSpatialFilter = 0x0
        nRepeatCount = 1
        bAllLayers = 0
        pszSQLStatement = 0x0
        pszDialect = 0x0
        poDS = 0x80518c0
        poDriver = 0x80514c0

and for the second is:

#0  0x41e23d7a in OGRGeoJSONReader::GenerateFeatureDefn (this=<value optimized out>, poObj=<value optimized out>) at ogrgeojsonreader.cpp:402
        poDefn = 0x80522a8
        bSuccess = <value optimized out>
        poObjProps = <value optimized out>
#1  0x41e23f39 in OGRGeoJSONReader::GenerateLayerDefn (this=<value optimized out>) at ogrgeojsonreader.cpp:319
        bSuccess = true
        objType = <value optimized out>
        poLayerDefn = <value optimized out>
#2  0x41e24adc in OGRGeoJSONReader::ReadLayer (this=<value optimized out>, pszName=<value optimized out>, poDS=<value optimized out>) at ogrgeojsonreader.cpp:116
        objType = <value optimized out>
        poSRS = <value optimized out>
#3  0x41e20fec in OGRGeoJSONDataSource::LoadLayer (this=<value optimized out>) at ogrgeojsondatasource.cpp:492
        err = <value optimized out>
        poLayer = <value optimized out>
        reader = Could not find the frame base for "OGRGeoJSONDataSource::LoadLayer()".

[Even Rouault]

Not sure why you got 2 different stack traces. The issue is the same.

Fixed in trunk (r18720) and in branches/1.7 (r18721). Tested in r18722.

[homme]

Replying to rouault:

Not sure why you got 2 different stack traces. The issue is the same.

Ah - I presumed because the traces were different so was the issue.

Fixed in trunk (r18720) and in branches/1.7 (r18721). Tested in r18722.

Thanks very much!