Member named `st_blocks' is missing from struct stat in sys/stat.h

[kaoutsis]

specs ​http://www.opengroup.org/onlinepubs/000095399/basedefs/sys/stat.h.html

[kaoutsis]

Required by wine:

/Haiku1/packages/wine-1.0-rc1/dlls/ntdll> make
gcc -c -I. -I. -I../../include -I../../include  -D__WINESRC__ -D_NTSYSTEM_ -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith  -g -O2  -o directory.o directory.c
/Haiku1/packages/wine-1.0-rc1/dlls/ntdll/directory.c: In function `append_entry':
/Haiku1/packages/wine-1.0-rc1/dlls/ntdll/directory.c:947: structure has no member named `st_blocks'
make: *** [directory.o] Error 1

directory.c at line 947 +-1:

        info->EndOfFile.QuadPart = st.st_size;
        info->AllocationSize.QuadPart = (ULONGLONG)st.st_blocks * 512;
        info->FileAttributes |= FILE_ATTRIBUTE_ARCHIVE;

[andreasf]

This has been a common issue for ports, see some examples and workarounds at HaikuPorts (formerly BePorts). BeOS did not have it either.

[nielx]

Git also assumes this st_blocks to exist.

[bonefish]

stat::st_blocks is actually part of the XSI extension.

[scottmc]

Replying to nielx:

Git also assumes this st_blocks to exist.

So does CVS 1-12-13...

[axeld]

Fixed in hrev27791. What unit is used for st_blocks in the software that it uses besides WINE which assumes a block size of 512?
Currently, BFS uses its native block size for the computation of the value, but if it's always expressed in 512 blocks, then we should probably change this.

Also note that this is just another value that cannot be retrieved via BStatable::GetStat() due to binary compatibility.

[anevilyak]

The POSIX spec seems to indicate that it's allowed to vary:

The unit for the st_blocks member of the stat structure is not defined within IEEE Std 1003.1-2001. In some implementations it is 512 bytes. It may differ on a file system basis. There is no correlation between values of the st_blocks and st_blksize, and the f_bsize (from <sys/statvfs.h>) structure members.

Source: ​http://www.opengroup.org/onlinepubs/000095399/basedefs/sys/stat.h.html

[axeld]

The POSIX spec is pretty stupid in this regard, as it makes the st_blocks field essentially useless. So if other platforms are using a consistent value, we should use that, too.

[anevilyak]

Right, I was wondering about that as well, since I don't see from the spec how you would actually determine the block size on a particular platform.