Support for DMS and UFS files > 2 GB

For processing file systems that contain files > 2 gigabytes (GB) a 64-bit variant exists for each of the following 32-bit C Library functions. The 64-bit functions differ from the corresponding 32-bit functions in that they have the suffix “64” in their names.

creat:	creat64
fgetpos:	fgetpos64
fopen:	fopen64
freopen:	freopen64
fseek:	fseek64
fseeko:	fseeko64
fsetpos:	fsetpos64
ftell:	ftell64
ftello:	ftello64
lseek:	lseek64
open:	open64
tmpfile:	tmpfile64

32-bit and 64-bit C/C++ library functions

There is no difference in terms of functionality between the 32-bit variant of a function and the associated 64-bit variant. The only differences concern the data types for parameters and return values if these specify an offset or a file position, since offset and return values > 2 GB must possible in order to process files > 2 GB. Thus, in addition to the 32-bit data type off_t, for example, there is also a 64-bit data type called off64_t.

The compilation environment makes available all the explicit 64-bit functions and types in addition to the 32-bit functions and types. A program can thus use either interface, as required.

• The 64-bit functions are only available with ANSI functionality.
• Since most of the names of the 64-bit functions are no longer unique CRTE-wide when truncated to 8 characters, sources that want to use 64-bit functions have to be generated as LLMs.

Using the 64-bit interface

The _FILE_OFFSET_BITS define allows you to choose between two alternatives for using the 64-bit interface:

• using 64-bit functions transparently (_FILE_OFFSET_BITS 64)

• calling 64-bit functions explicitly (_FILE_OFFSET_BITS 32)

• The _FILE_OFFSET_BITS define must be set on an include file before the first include.
• You can replace 32-bit functions with 64-bit functions automatically by means of name defines or macro defines (see section “Preprocessor define_MAP_NAME”).

Using 64-bit functions transparently (_FILE_OFFSET_BITS 64)

The _FILE_OFFSET_BITS 64 define allows the 64-bit interface to be used transparently, since the 32-bit functions contained in the source code are automatically replaced with the associated 64-bit variants during compilation (with the exception of fseek and ftell, see below). In addition, the compilation environment makes data types available in the appropriate size. The data type off_t, for example, is declared as long long.

You can use the _MAP_NAME preprocessor define to specify whether the 32-bit functions are to be mapped to 64-bit functions by means of the name define method or the macro define method (see "Preprocessor define _MAP_NAME").

A program can process both files > 2 GB and files <= 2 GB. Transparent use of the 64-bit functions permits programs that were previously designed only for files <= 2 GB to process files > 2GB without the need for any changes to the source code.

The functions fseek and ftell cannot be automatically replaced with fseek64 and ftell64. Please use the functions fseeko and ftello if you want automatic replacement to be carried out.

Calling 64-bit functions explicitly

If the _FILE_OFFSET_BITS 32 define is set or if _FILE_OFFSET_BITS is not defined, you have to use the 64-bit variants of the file processing functions described above in order to process files > 2 GB:

• If you try to process a file > 2 GB using a 32-bit variant, this leads to abortion.

• If you use the 64-bit variants, however, you can also process files <= 2 GB.

You can only use the 64-bit functions explicitly if the _LARGEFILE64_SOURCE 1 define is set beforehand (prototype generation and further defines).