threads / patch / 2952

patchdiff-tree: stop on broken output pipe

Subject: [PATCH] diff-tree: stop on broken output pipe

## tl;dr

5 messages between Jan 2, 2006 and Jan 7, 2006. Diffs are folded; open one to read it.

replies: 4people: 3as markdown or json

Johannes Schindelin· Jan 2, 2006, 00:17 UTC · lore

---

	Without this, on my iBook git-whatchanged keeps running when I 
	quit "less". I have to interrupt the process a second time. No
	idea why it works on Linux.

 diff-tree.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

a3ff3aa99f1cc7d1e59563fbf22fed8a4f3f2d24

Show changes to diff-tree.c +3 −2

diff --git a/diff-tree.c b/diff-tree.c
index efa2b94..c06ccd3 100644
--- a/diff-tree.c
+++ b/diff-tree.c
@@ -25,8 +25,9 @@ static int call_diff_flush(void)
 		return 0;
 	}
 	if (header) {
-		if (!no_commit_id)
-			printf("%s%c", header, diff_options.line_termination);
+		if (!no_commit_id && printf("%s%c", header,
+				diff_options.line_termination) < 0)
+			die("broken output pipe");
 		header = NULL;
 	}
 	diff_flush(&diff_options);

-- 
1.0.GIT

Andreas Ericsson· Jan 2, 2006, 01:21 UTC · re: Johannes Schindelin · lore

Re: [PATCH] diff-tree: stop on broken output pipe

Johannes Schindelin wrote:

Show 6 quoted lines

> ---
> 
> 	Without this, on my iBook git-whatchanged keeps running when I 
> 	quit "less". I have to interrupt the process a second time. No
> 	idea why it works on Linux.
>

On Linux the sending end dies when it catches SIGPIPE. I would have thought that should happen on OSX too. What shell are you running?

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Johannes Schindelin· Jan 2, 2006, 11:01 UTC · re: Andreas Ericsson · lore

Re: [PATCH] diff-tree: stop on broken output pipe

Hi,

On Mon, 2 Jan 2006, Andreas Ericsson wrote:

Show 10 quoted lines

> Johannes Schindelin wrote:
> > ---
> > 
> > 	Without this, on my iBook git-whatchanged keeps running when I 	quit
> > "less". I have to interrupt the process a second time. No
> > 	idea why it works on Linux.
> > 
> 
> On Linux the sending end dies when it catches SIGPIPE. I would have thought
> that should happen on OSX too. What shell are you running?

Why, bash 2.05a0(1)-release (powerpc-app,e-darwin6.0). Is there a way to turn off SIGPIPEs? Could be turned off by mistake here...

Ciao, Dscho

Junio C Hamano· Jan 6, 2006, 22:36 UTC · re: Johannes Schindelin · lore

Re: [PATCH] diff-tree: stop on broken output pipe

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Show 10 quoted lines

> 	Without this, on my iBook git-whatchanged keeps running when I 
> 	quit "less". I have to interrupt the process a second time. No
> 	idea why it works on Linux.
>
> -		if (!no_commit_id)
> -			printf("%s%c", header, diff_options.line_termination);
> +		if (!no_commit_id && printf("%s%c", header,
> +				diff_options.line_termination) < 0)
> +			die("broken output pipe");
>  		header = NULL;

There are tons of output other than this printf() and catching an output error only on this one and assuming that is a broken pipe somehow feels doubly wrong to me.

Johannes Schindelin· Jan 7, 2006, 00:41 UTC · re: Junio C Hamano · lore

Re: [PATCH] diff-tree: stop on broken output pipe

Hi,

On Fri, 6 Jan 2006, Junio C Hamano wrote:

Show 16 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> > 	Without this, on my iBook git-whatchanged keeps running when I 
> > 	quit "less". I have to interrupt the process a second time. No
> > 	idea why it works on Linux.
> >
> > -		if (!no_commit_id)
> > -			printf("%s%c", header, diff_options.line_termination);
> > +		if (!no_commit_id && printf("%s%c", header,
> > +				diff_options.line_termination) < 0)
> > +			die("broken output pipe");
> >  		header = NULL;
> 
> There are tons of output other than this printf() and catching
> an output error only on this one and assuming that is a broken
> pipe somehow feels doubly wrong to me.

Welcome back! Hope you had a good time, and the fish tasted fine.

There really are tons of printf(), but

- I did not want to be invasive,
- if one printf() fails, the next will, too (probably), and
- the problem occurs only on macosx here.

Maybe it is not correct to assume that the output pipe is broken, but it is an error to keep going when printf() fails. So I picked one printf() which is called regularly.

Andreas said that a SIGPIPE should be thrown, and bash should kill the script. I have not found out why my bash doesn't. Until I do, I will have this patch in my local git (and a similar patch in git-rev-list). It just is plain annoying when git-whatchanged (the git script I use most often) hangs there, burning cycles.

Ciao, Dscho