Closed pokulo closed 10 years ago
As far as I understood returning the _inputcontainerconnector
is intended behaviour, as a connector should be able to implement both reading and writing. This is documented somewhere, but in this case naming leads to wrong interpretation quickly which I think should be tackled.
However, if the connector decides to overwrite data a log statement wouldn't help a lot by letting the user know that data has been overwritten.
I suggest that if there is a R/W connector, the object/pointer should be set to _inputcontainerconnector
and _outputcontainerconnector
. Would this solve the problem?
ConnectTo now has a boolean return value.
if a call of a
connectTo(..)
fails (e.g. bacause of wrong parameters), an ERROR is emmited, but no return value (void) reports that to the programmer. So I cannot check wether callingstore()
should be called.The lack of a status reporting return value was not the first initial error causing this bugreport. A rather general issue occurs on the current behaviour of the
containerconnector()
method, which returns the_inputcontainerconnector
if there is no_outcontainerconnector
without even reporting to log. This may cause overwriting input data with manipulated without notice of the user and thus shouldn't be possible. Instead rather report an error and handlereturn NULL
;My
store()
call caused the gdalConnector to crash (SIGSEGV) on a call like this:line125
Because the raster I wanted to store was created by rastercalculation and this way had not even an
_incontainerconnector
.