[Gdal-dev] OGR PostGIS driver problems on x86_64

[Edzer J. Pebesma]

Hi,

I'm running debian testing ("etch") on a x86_64 architecture; when I run

ogr2ogr -overwrite -f "PostgreSQL" "PG:dbname=postgis" meuse.shp

it gives me a segmentation fault. A stack trace is

0x00002aaacf4b25b0 in strlen () from /lib/libc.so.6
(gdb) bt
#0  0x00002aaacf4b25b0 in strlen () from /lib/libc.so.6
#1  0x00002aaacf4844bc in vfprintf () from /lib/libc.so.6
#2  0x00002aaacf4a472a in vsnprintf () from /lib/libc.so.6
#3  0x00002aaacd6af5e5 in CPLString::vPrintf () from
/usr/lib/libgdal1.3.2.so.1
#4  0x00002aaacd6af80c in CPLString::Printf () from
/usr/lib/libgdal1.3.2.so.1
#5  0x00002aaacd7565d7 in OGRPGTableLayer::CreateFeatureViaInsert ()
from /usr/lib/libgdal1.3.2.so.1
#6  0x0000000000401b83 in ?? ()
#7  0x00000000004023f5 in ?? ()
#8  0x00002aaacf45c4ca in __libc_start_main () from /lib/libc.so.6
#9  0x00000000004017aa in ?? ()
#10 0x00007fffdd741158 in ?? ()
#11 0x0000000000000000 in ?? ()

when I repeat this on an otherwise identical i686 (32 bits)
architecture, no problems. I found this out when trying to write PostGIS 
data directly from R, using the rgdal interface.

Is this a known problem? How can I (help to) resolve this?

A related but less serious issue is that the layer name is ignored: on 
the 32 bits architecture

ogr2ogr -overwrite -f "PostgreSQL" "PG:dbname=postgis" meuse.shp mmm

will not add a table; after removing mmm table meuse will be written,
after renaming the shapefile into mmm.shp (etc) mmm will be written. A
layer name different from the shapefile name causes it not to be
written; no layer name will result in the shapefile name being written. 
Is this documented behaviour?
--
Edzer