All of lore.kernel.org
 help / color / mirror / Atom feed
From: Junio C Hamano <gitster@pobox.com>
To: skimo@liacs.nl
Cc: git@vger.kernel.org, Steffen Prohaska <prohaska@zib.de>,
	Johannes.Schindelin@gmx.de
Subject: Re: [PATCH v4] git-apply: apply submodule changes
Date: Tue, 14 Aug 2007 13:00:36 -0700	[thread overview]
Message-ID: <7vmywtc2e3.fsf@assigned-by-dhcp.cox.net> (raw)
In-Reply-To: <20070813171349.GL999MdfPADPa@greensroom.kotnet.org> (Sven Verdoolaege's message of "Mon, 13 Aug 2007 19:13:49 +0200")

Sven Verdoolaege <skimo@kotnet.org> writes:

> @@ -2096,8 +2142,15 @@ static int check_patch(struct patch *patch, struct patch *prev_patch)
>  				    lstat(old_name, &st))
>  					return -1;
>  			}
> -			if (!cached)
> +			if (!cached) {
>  				changed = ce_match_stat(ce, &st, 1);
> +				if (S_ISGITLINK(patch->old_mode)) {
> +					changed &= TYPE_CHANGED;
> +					if (!changed &&
> +					    verify_gitlink_clean(patch->old_name))
> +						changed |= TYPE_CHANGED;
> +				}
> +			}

In this codepath, we know the patch wants to either modify the
path at old_name or remove old_name.  If we are going to affect
the work tree, we have run lstat on it, and ran checkout_entry() 
if we did not have anything there and did lstat() again.

I think the check "S_ISGITLINK(patch->old_mode)" is wrong
(that's where my confusion while reading your patch came from).
It has to check ce's mode, not patch->old_mode, because we are
verifying if the index matches with the work tree in this
codepath.  If you fix it to S_ISGITLINK(ntohl(ce->ce_mode)),
I think I can see what you are trying to do.

When ce is not a gitlink, you keep the original behaviour, which
is assuring that you did not break things for people who do not
use gitlink.

I am still having trouble with the TYPE_CHANGED bits.  You
discard everything other than TYPE_CHANGED, and 

 - if ce_match_stat() returned TYPE_CHANGED, then that is given
   to later processing to cause us to fail "oops, path is not up
   to date";

 - if ce_match_stat() did not return TYPE_CHANGED, that means we
   found a directory at the path (ce_match_stat_basic() says
   so).  In such a case you call verify_gitlink_clean(), but it
   essentially says "make sure there is either an empty
   directory or some repository".  Maybe we do not even have to
   have this extra check?

When ce is a gitlink, ce_match_stat() says DATA_CHANGED if the
commit in the work tree of the subproject is different.  From
the earlier discussions, we do want to discard DATA_CHANGED for
this codepath.

So it looks almost Ok after spending a few days looking at this
code.  Finally.

However, if it takes _me_ three days to understand this hunk,
(admittably, the parameter to S_ISGITLINK() completely confused
me originally, and I also had other things to do, so it was not
"72 hours"), I do not think the code with your patch is
maintainable by anybody.  At least we would need to have a few
words of comment to describe what is going on there.

	if (!cached) {
        	changed = ce_match_stat(ce, &st, 1);
                if (S_ISGITLINK(ntohl(ce->ce_mode)))
                	/*
			 * ce_match_stat() reports the
			 * difference between the commit object
                         * name in the index and what is checked
			 * out in the work tree of subproject;
                         * because we do not recurse, we do not
			 * want to insist on them matching with
                         * each other.
                         */
                	changed &= ~DATA_CHANGED;
	}
        if (changed)
        	return error("%s: does not match index", old_name);

      parent reply	other threads:[~2007-08-14 20:00 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-08-10  9:30 [PATCH] git-apply: apply submodule changes Sven Verdoolaege
2007-08-10 12:34 ` Johannes Schindelin
2007-08-10 12:39 ` Johannes Schindelin
2007-08-10 13:57 ` [PATCH resend] " Sven Verdoolaege
2007-08-11  5:43   ` Junio C Hamano
2007-08-11  6:45     ` Sven Verdoolaege
2007-08-11  7:00       ` Junio C Hamano
2007-08-12 14:23 ` [PATCH v3] " Sven Verdoolaege
2007-08-12 18:16   ` Junio C Hamano
2007-08-12 18:50     ` Sven Verdoolaege
2007-08-12 19:24       ` Junio C Hamano
2007-08-13  9:37         ` Sven Verdoolaege
2007-08-13 17:13           ` [PATCH v4] " Sven Verdoolaege
2007-08-13 20:26             ` Junio C Hamano
     [not found]               ` <7vd4xqeilh.fsf@assigned-by-dhcp.cox.net>
2007-08-14  8:39                 ` Sven Verdoolaege
2007-08-14  9:09                   ` Junio C Hamano
2007-08-15 17:22                     ` [PATCH v6] " Sven Verdoolaege
2007-08-16  0:02                       ` Junio C Hamano
2007-08-14 20:00             ` Junio C Hamano [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=7vmywtc2e3.fsf@assigned-by-dhcp.cox.net \
    --to=gitster@pobox.com \
    --cc=Johannes.Schindelin@gmx.de \
    --cc=git@vger.kernel.org \
    --cc=prohaska@zib.de \
    --cc=skimo@liacs.nl \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.