#187 svn:executable property on Windows filesystems does have an effect

[Anonymous]

Originally created by: spt5...@gmail.com

See Advanced Topics->File Portability->File Executability

This property has no effect on filesystems that have no concept of an executable permission bit, such as FAT32 and NTFS.[17]

There is, indeed, a notion of executable permission bit on NTFS. I found this out the hard way. I suggest the following rewording:

This property has no effect on filesystems that have no concept of an executable permission bit, such as FAT32.[17]

[Anonymous]

Originally posted by: cmpilato

(No comment was entered for this change.)

Labels: Milestone-whenever

[Daniel Sahlberg]

I'm not so sure about this.

Indeed NTFS has a "Read and execute" permission, but Subversion doesn't (as far as I can tell) sent any NTFS permissions.

I tried to trace the svn:executable property and it seems to end up in svn_io_set_file_executable. There is a comment:
[[[ On Windows and OS/2, just exit -- on unix call our internal function which attempts to honor the umask. ]]]

According to my tests, a new file inherits the "Read and execute" permission from the parent directory when it is checked out, no matter the svn:executable property.

I'm suggesting to close this ticket.

[C. Michael Pilato]

As I read this, I believe the OP is not suggesting that Subversion behaves any differently than the book describes, but rather that the text misrepresents NTFS as not having a notion of an executable permission bit. Unfortunately, the proposed fix causes a different confusion by allowing an NTFS-knowledgeable person to assume that Subversion will conditionally set that bit on NTFS (which, of course, it does not).

Perhaps the text could be changed to read instead:

This property has no effect on Windows or on filesystems that have no concept of an executable permission bit (such as FAT32).

?

[Daniel Sahlberg]

(y)
It also doesn't have an effect on OS2, but I don't know how relevant it is these days.

[C. Michael Pilato]

I still have a lingering nag here. I assumed the executable bit was a feature of the filesystem, not of the OS. Yet, "system calls" are a feature of the OS.

It's common to have, say, a FAT32 volume mounted in Linux. But does the reverse occur, where an ext3 filesystem is mounted in Windows? It sounds like APR won't try to set an executable bit on such a filesystem.

What about network filesystems? Will APR preserve the executable bit on, say, an NFS mount when accessed via Windows?