git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jeff King <peff@peff.net>
To: Augie Fackler <augie@google.com>
Cc: Junio C Hamano <gitster@pobox.com>, Johannes Sixt <j6t@kdbg.org>,
	git@vger.kernel.org, Stefan Beller <sbeller@google.com>
Subject: Re: [PATCH 3/3] pkt-line: support tracing verbatim pack contents
Date: Tue, 16 Jun 2015 13:10:06 -0400	[thread overview]
Message-ID: <20150616171005.GB18667@peff.net> (raw)
In-Reply-To: <CAHcr6HYvVR4uTmtegWHK0h+v_aVs4JVLsSwvjthGY3pb=-Q0yQ@mail.gmail.com>

On Tue, Jun 16, 2015 at 11:38:41AM -0400, Augie Fackler wrote:

> > We already have a GIT_TRACE_PACKET mechanism for tracing
> > packets. Let's extend it with GIT_TRACE_PACK to dump the
> > verbatim packfile.
> 
> FWIW, this also works for me - I have no preference between my patches
> and Jeff's. I suspect yours are much better given that you have a clue
> about git internals ;).

I like mine better than yours, if only because it hooks into the
existing tracing mechanism. But I am sad that neither of our proposals
works for tracing pushed packs (something that is useful for the
opposite situation: you have a sane git server, and some unknown
misbehaving _client_ is sending you bogus packs).

Here's a rough cut at the "trace stdin" idea I mentioned earlier (which
is essentially an internal "tee"). You can collect the incoming pack
like:

  GIT_TRACE=1 \
  GIT_TRACE_STDIN=/tmp/foo.pack \
  GIT_TRACE_STDIN_FOR=index-pack \
  git fetch ...

The "STDIN_FOR" hack is there because otherwise you get the stdin for
multiple processes intermingled. I think a cleaner way to do it would be
to allow something like:

  GIT_TRACE_STDIN=/tmp/stdin.%p

and replace "%p" with the PID of the tracing process.

I like this approach because it works everywhere (for pushed packs, but
also for any other git commands you may want to debug).  The downside
versus the other patches is that to replay the index-pack command, you
need to collect both its stdin and its arguments (which you can get from
the GIT_TRACE output). On the other hand, it gives you a more realistic
replay of what happened (e.g., if there is a bug triggered by the
--pack-header code).

Here is the patch:

---
diff --git a/git.c b/git.c
index 44374b1..d58866e 100644
--- a/git.c
+++ b/git.c
@@ -616,6 +616,80 @@ static void restore_sigpipe_to_default(void)
 	signal(SIGPIPE, SIG_DFL);
 }
 
+static int copy_stdin(int in, int out, void *data)
+{
+	struct trace_key *key = data;
+	while (1) {
+		char buf[8192];
+		ssize_t len = xread(in, buf, sizeof(buf));
+		if (!len)
+			break;
+		if (len < 0) {
+			warning("error reading stdin trace: %s",
+				strerror(errno));
+			break;
+		}
+
+		trace_verbatim(key, buf, len);
+		if (write_in_full(out, buf, len) < 0) {
+			warning("error writing stdin trace: %s",
+				strerror(errno));
+			break;
+		}
+	}
+	close(in);
+	close(out);
+	return 0;
+}
+
+static int trace_stdin_for(const char *cmd)
+{
+	int ret = 1;
+	const char *x = getenv("GIT_TRACE_STDIN_FOR");
+
+	if (x) {
+		struct string_list items = STRING_LIST_INIT_DUP;
+		string_list_split(&items, x, ':', -1);
+		ret = unsorted_string_list_has_string(&items, cmd);
+		string_list_clear(&items, 0);
+	}
+
+	return ret;
+}
+
+static void trace_stdin(const char *cmd)
+{
+	static struct trace_key key = TRACE_KEY_INIT(STDIN);
+	static struct async async;
+
+	if (!trace_stdin_for(cmd) || !trace_want(&key))
+		return;
+
+	memset(&async, 0, sizeof(async));
+	async.proc = copy_stdin;
+	async.data = &key;
+	async.in = dup(0);
+	async.out = -1;
+
+	if (async.in < 0 || start_async(&async) < 0) {
+		warning("unable to trace stdin: %s", strerror(errno));
+		return ;
+	}
+
+	/*
+	 * At this point we've handed stdin off to the async process,
+	 * so there we are past the point of no return.
+	 */
+	if (dup2(async.out, 0))
+		die_errno("unable to redirect stdin from async process");
+	close(async.out);
+
+	/*
+	 * leak async; we would know to finish_async() only when we are
+	 * exiting, and there is no point then
+	 */
+}
+
 int main(int argc, char **av)
 {
 	const char **argv = (const char **) av;
@@ -674,6 +748,8 @@ int main(int argc, char **av)
 	}
 	cmd = argv[0];
 
+	trace_stdin(cmd);
+
 	/*
 	 * We use PATH to find git commands, but we prepend some higher
 	 * precedence paths: the "--exec-path" option, the GIT_EXEC_PATH

  parent reply	other threads:[~2015-06-16 17:10 UTC|newest]

Thread overview: 33+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <CAGZ79kaS4utvDbXOo7emmSUH6M-8LY-oA65Ss3PLDkFModkbSg@mail.gmail.com>
2015-06-11 18:59 ` [PATCH v2] fetch-pack: optionally save packs to disk Augie Fackler
2015-06-12  6:22   ` Johannes Sixt
2015-06-12 15:07     ` Junio C Hamano
2015-06-12 17:02       ` Augie Fackler
2015-06-12 18:00       ` Jeff King
2015-06-12 21:25         ` Jeff King
2015-06-12 21:28           ` [PATCH 1/3] pkt-line: simplify starts_with checks in packet tracing Jeff King
2015-06-12 21:35             ` Stefan Beller
2015-06-12 21:28           ` [PATCH 2/3] pkt-line: tighten sideband PACK check when tracing Jeff King
2015-06-12 21:39             ` Stefan Beller
2015-06-12 21:41               ` Jeff King
2015-06-12 21:43                 ` Stefan Beller
2015-06-12 21:28           ` [PATCH 3/3] pkt-line: support tracing verbatim pack contents Jeff King
2015-06-16 15:38             ` Augie Fackler
2015-06-16 16:39               ` Junio C Hamano
2015-06-16 16:43                 ` Jeff King
2015-06-16 16:52                   ` Augie Fackler
2015-06-16 17:23                     ` Jeff King
2015-06-16 17:10               ` Jeff King [this message]
2015-06-16 17:14                 ` Augie Fackler
2015-06-16 17:18                   ` Jeff King
2015-06-16 17:23                     ` Augie Fackler
2015-06-16 19:31                       ` [PATCH/RFC 0/3] add GIT_TRACE_STDIN Jeff King
2015-06-16 19:35                         ` [PATCH 1/3] trace: implement %p placeholder for filenames Jeff King
2015-06-16 19:36                         ` [PATCH 2/3] trace: add pid to each output line Jeff King
2015-06-16 19:37                         ` [PATCH 3/3] trace: add GIT_TRACE_STDIN Jeff King
2015-06-16 19:49                           ` Jeff King
2015-06-16 21:20                             ` Jeff King
2015-06-17 10:04                               ` Duy Nguyen
2015-06-17 19:10                                 ` Jeff King
2015-06-18 10:20                                   ` Duy Nguyen
2015-06-26 18:47                                   ` Junio C Hamano
2015-06-12 16:54     ` [PATCH v2] fetch-pack: optionally save packs to disk Augie Fackler

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=20150616171005.GB18667@peff.net \
    --to=peff@peff.net \
    --cc=augie@google.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=j6t@kdbg.org \
    --cc=sbeller@google.com \
    /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).