XrdOss: implement proposed solution for better rename POSIX compliancy

[gganis]

Implements enhancement #659

[abh3]

What this means is that any symlinked FS will not allow a rename on top of an existing link. Is that what you really want? That excludes any virtualized filesystem even if local.

[gganis]

The symlink part comes from your original comment. In my case I am mostly interested to local file, no symlinks. However, I see that the man page of rename says that

if newpath refers to a symbolic link, the link will be overwritten

So I suggest that we only leave the check on 'remotefs'. What do you think?

[abh3]

Well, unfortuantely, in a symlinked FS that is not a simple operation. The symlink refers to a data file and the data file has an xattr that records its symlink name. All of that has to change to successfully overwrite the symlink. Merely removing the symlink test would really screw things up for many people. I guess that's why we forbid renaming on top of an existing file. I don't have a problem of restricting destructive renames to simple file systems as long as that's OK with you. Andy

…

On Thu, 1 Mar 2018, Gerardo GANIS wrote: gganis commented on this pull request. > // -#if 0 - retc2 = lstat(local_path_New, &statbuff); - if (!retc2) return -EEXIST; -#endif + if (!(retc2 = lstat(local_path_New, &statbuff))) + { if (remotefs || (statbuff.st_mode & S_IFMT) == S_IFLNK) return -EEXIST; + } The symlink part comes from your original comment. In my case I am mostly interested to local file, no symlinks. However, I see that the man page of rename says that if newpath refers to a symbolic link, the link will be overwritten So I suggest that we only leave the check on 'remotefs'. What do you think? -- You are receiving this because you commented. Reply to this email directly or view it on GitHub: #660 (comment)

[gganis]

I don't have a problem of restricting destructive renames to simple
file systems as long as that's OK with you.

That's ok with me. Does it mean that the PR can stay as it is?
Gerri

[abh3]

Yes, I'll merge it. Andy

…

On Thu, 1 Mar 2018, Gerardo GANIS wrote: > I don't have a problem of restricting destructive renames to simple file systems as long as that's OK with you. Tha's ok with me. Does it mean that the PR can stay as it is? Gerri -- You are receiving this because you commented. Reply to this email directly or view it on GitHub: #660 (comment)