[gdal-dev] Disk or database full error from Rasterlite driver

[a.furieri at lqt.it]

On Wed, 9 May 2012 19:50:15 +0000 (UTC), Jukka Rahkonen wrote:
> However, the Rasterlite file size has probably been
> around 4 gigabytes when the error occurred. Did I meet some file size
> limit or tile count limit of the Rasterlite driver?
>

Hi Jukka,

on any platform there is an intrinsic physical size limit;
this is simply due to the word size measured in bits.
on 32 bit platforms the max address you can express is 2^32,
i.e. 4 GB; as you probably remember, on earlier versions of
Linux and Win98 this was the max file-size allowed by their
file-systems.

Happily enough, any modern 32 bit Linux supports a file-system
based on long-addressing, i.e. using two 32 bit words in order
to support 64 bit file offsets; and this effectively allows to
store files widely exceeding the 4 GB barrier.

More or less the same is on WinXP, WinVista and Win7 if you are
using the NTFS file-system.
But if you are using instead the obsolete FAT file-system (the
default choice on many USB pen drivers and USB external HDDs) the
old 32 bit addressing is the only one supported, thus still imposing
a file-size limit of 4 GB.

A final consideration: on both Linux 32 bit and Windows 32 bit you
must explicitly set few compile-time flags in order to build an
executable effectively supporting 64 bit extended file addressing;
on both GCC and MSVC these compiler options are:

-D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE=1 -D_LARGE_FILE=1

conclusion: SQLite, SpatiaLite and RasterLite have no intrinsic
file-size limits (I've successfully built several impressively huge
DB-files 100+ GB).
- using 64 bit sw this assumption is surely true
- using 32 bit sw it depends: you are surely limited to 4 GB max,
   if your binaries were built without setting the above mentioned 
flags.
- if you are using the infamous FAT file-system there is absolutely no
   way to escape from the 4 GB barrier because this is the physical 
limit
   imposed by this file-system.

bye Sandro


-- 
Il messaggio e' stato analizzato alla ricerca di virus o
contenuti pericolosi da MailScanner, ed e'
risultato non infetto.