Re: [PATCH] fix compare symlink against readlink not data

[Junio C Hamano <junkio@cox.net>]

>>>>> "GKH" == Greg KH <greg@kroah.com> writes:

GKH> On Fri, May 06, 2005 at 03:45:01PM +0200, Kay Sievers wrote:
>> Thanks to Greg for recognizing this while he added our list of symlinks
>> back to the udev repository.

GKH> Hm, even with this patch applied (it's in Linus's tree right now), I
GKH> still get the following with a clean checked out udev tree:
GKH>  $ cg-diff
GKH>  Index: test/sys/block/cciss!c0d0/device
GKH>  ===================================================================
GKH>  Index: test/sys/block/rd!c0d0/device
GKH>  ===================================================================

GKH> Any idea?

I do not use Cogito but probably it is this piece of code in
cg-Xdiffdo.  It is assuming that a valid SHA1 means it can diff
against the filesystem object that resides there.  It used to be
the case before symlinks but not anymore.

    mkbanner () {
            loc=$1; treeid=$2; fname=$3; mode=$4; sha1=$5;

            if [ "$sha1" != "0000000000000000000000000000000000000000" ]; then
                    git-cat-file blob $sha1 >$loc
            else
                    ln -s "$(pwd)/$fname" "$loc"
                    sha1="!"
            fi

Maybe changing the if to (I'm writing this in e-mail editor so
completely untested) something like:

    if expr "$sha1" : '0*$' >/dev/null ||
       expr "$mode" : '.*120000$' >/dev/null
    then
        git-cat-file blob "$sha1" >$loc

would help?

Also could you try the low-level git command, git-diff-cache -p,
against the tree you are comparing?  The built-in diff stuff
might get this wrong too.