From: Johannes Sixt <j6t@kdbg.org>
Date: Mon, 12 Sep 2016 08:35:22 GMT
Subject: Re: [PATCH v2 21/25] sequencer: refactor write_message()
Message-ID: <e21e3391-bfc3-68b8-8eaf-0e15e5a436c5@kdbg.org>
In-Reply-To: <da2293aee439da2274e30304e2d1f097b9644e64.1473590966.git.johannes.schindelin@gmx.de>

Am 11.09.2016 um 12:55 schrieb Johannes Schindelin:
> -static int write_message(struct strbuf *msgbuf, const char *filename)
> +static int write_with_lock_file(const char *filename,
> +				const void *buf, size_t len, int append_eol)
>  {
>  	static struct lock_file msg_file;
>
>  	int msg_fd = hold_lock_file_for_update(&msg_file, filename, 0);
>  	if (msg_fd < 0)
>  		return error_errno(_("Could not lock '%s'"), filename);
> -	if (write_in_full(msg_fd, msgbuf->buf, msgbuf->len) < 0)
> -		return error_errno(_("Could not write to %s"), filename);
> -	strbuf_release(msgbuf);
> +	if (write_in_full(msg_fd, buf, len) < 0)
> +		return error_errno(_("Could not write to '%s'"), filename);
> +	if (append_eol && write(msg_fd, "\n", 1) < 0)
> +		return error_errno(_("Could not write eol to '%s"), filename);
>  	if (commit_lock_file(&msg_file) < 0)
>  		return error(_("Error wrapping up %s."), filename);
>
>  	return 0;
>  }

The two error paths in the added lines should both

		rollback_lock_file(&msg_file);

, I think. But I do notice that this is not exactly new, so...

-- Hannes