Re: pg_tablespace_location() failure with allow_in_place_tablespaces

On Tue, Mar 8, 2022 at 12:58 AM Michael Paquier <michael@paquier.xyz> wrote:
> Junction points are directories, no?  Are you sure that this works
> correctly on WIN32?  It seems to me that we'd better use readlink()
> only for entries in pg_tlbspc/ that are PGFILETYPE_LNK on non-WIN32
> and pgwin32_is_junction() on WIN32.

Hmm.  So the code we finished up with in the tree looks like this:

#ifdef WIN32
            if (!pgwin32_is_junction(fullpath))
                continue;
#else
            if (get_dirent_type(fullpath, de, false, ERROR) != PGFILETYPE_LNK)
                continue;
#endif

As mentioned, I was unhappy with the lack of error checking for that
interface, and I've started a new thread about that, but then I
started wondering if we missed a trick here: get_dirent_type() contain
code that wants to return PGFILETYPE_LNK for reparse points.  Clearly
it's not working, based on results reported in this thread.  Is that
explained by your comment above, "junction points _are_ directories",
and we're testing the attribute flags in the wrong order here?

    if ((fd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) != 0)
        d->ret.d_type = DT_DIR;
    /* For reparse points dwReserved0 field will contain the ReparseTag */
    else if ((fd.dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) != 0 &&
             (fd.dwReserved0 == IO_REPARSE_TAG_MOUNT_POINT))
        d->ret.d_type = DT_LNK;
    else
        d->ret.d_type = DT_REG;