From: Jeff King <peff@peff.net>
To: git@vger.kernel.org
Cc: Junio C Hamano <gitster@pobox.com>,
Dave Borowitz <dborowitz@google.com>,
Shawn Pearce <spearce@spearce.org>
Subject: [PATCH] pack-protocol: document newline behavior in push commands
Date: Thu, 3 Sep 2015 04:02:05 -0400 [thread overview]
Message-ID: <20150903080205.GA20182@sigill.intra.peff.net> (raw)
Our pack-protocol spec indicates that a pushing client
should send ref update commands like:
$old_sha1 $new_sha1 $ref\n
with each ref update in its own pktline, with a trailing
newline. However, git itself does not follow this behavior;
it omits the trailing newline.
For the most part the distinction is harmless. The
`receive-pack` on the other end will call
`packet_read_line`, which strips off the newline if it is
there, and we are fine either way.
Things are more complicated for the initial ref, which also
has capabilities. The spec says to send:
$old_sha1 $new_sha1 $ref\0 $caps\n
which is also OK by the current `receive-pack code (the
newline is at the end of the packet, so we still strip it).
As an aside, it would _not_ be OK to send:
$old_sha1 $new_sha1 $ref\n\0 $caps\n
The spec does not allow that and receive-pack would reject
it, but it is perhaps a mistake that a naive client
implementation might make.
So what is in the current spec is quite reasonable, and
simply follows the normal rules for pkt-line framing (we say
LF, but it really is optional). But it does not document
how git actually behaves.
Signed-off-by: Jeff King <peff@peff.net>
---
+cc Junio, Dave, and Shawn, as this is somewhat related to the
discussion in
http://thread.gmane.org/gmane.comp.version-control.git/273175/focus=273444
I happened to be looking into some protocol stuff recently and noticed
that we do not follow the spec here (but interestingly, libgit2 does!).
Frankly, this feels a bit like a step backwards to me. I am tempted to
suggest instead that git start sending the newlines, but I'm not sure
it's worth any potential fallout.
I'm also tempted to scrap this and say it just falls under the rule
that every PKT-LINE is "sender SHOULD include LF" and "receiver MUST NOT
complain about missing LF" (which does appear earlier in the document,
though in a different context). Or maybe we should write out that rule
explicitly and drop the "LF" from all parts of the grammar (which is
what the thread above advocates).
Documentation/technical/pack-protocol.txt | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
index 4064fc7..9ce53b4 100644
--- a/Documentation/technical/pack-protocol.txt
+++ b/Documentation/technical/pack-protocol.txt
@@ -469,8 +469,8 @@ references.
shallow = PKT-LINE("shallow" SP obj-id LF)
- command-list = PKT-LINE(command NUL capability-list LF)
- *PKT-LINE(command LF)
+ command-list = PKT-LINE(command NUL capability-list)
+ *PKT-LINE(command)
flush-pkt
command = create / delete / update
@@ -586,8 +586,8 @@ An example client/server communication might look like this:
S: 003f74730d410fcb6603ace96f1dc55ea6196122532d refs/heads/team\n
S: 0000
- C: 003e7d1665144a3a975c05f1f43902ddaf084e784dbe 74730d410fcb6603ace96f1dc55ea6196122532d refs/heads/debug\n
- C: 003e74730d410fcb6603ace96f1dc55ea6196122532d 5a3f6be755bbb7deae50065988cbfa1ffa9ab68a refs/heads/master\n
+ C: 003e7d1665144a3a975c05f1f43902ddaf084e784dbe 74730d410fcb6603ace96f1dc55ea6196122532d refs/heads/debug
+ C: 003e74730d410fcb6603ace96f1dc55ea6196122532d 5a3f6be755bbb7deae50065988cbfa1ffa9ab68a refs/heads/master
C: 0000
C: [PACKDATA]
--
2.5.1.812.ge796bff
next reply other threads:[~2015-09-03 8:02 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-09-03 8:02 Jeff King [this message]
2015-09-03 8:24 ` [PATCH] pack-protocol: clarify LF-handling in PKT-LINE() Jeff King
2015-09-03 8:26 ` Jeff King
2015-09-03 22:19 ` Junio C Hamano
2015-09-03 22:17 ` [PATCH] pack-protocol: document newline behavior in push commands Junio C Hamano
2015-09-04 9:00 ` Jeff King
2015-09-04 15:49 ` Junio C Hamano
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=20150903080205.GA20182@sigill.intra.peff.net \
--to=peff@peff.net \
--cc=dborowitz@google.com \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=spearce@spearce.org \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).