Re: [RFC/PATCH] imap-send: Code correctness flagged by clang

From: Jonathan Nieder <jrnieder@gmail.com>
Date: Aug 7, 2010, 21:04 UTC
Message-ID: <20100807210429.GA2216@burratino>
In-Reply-To: <1281183136-10352-1-git-send-email-avarab@gmail.com>

Ævar Arnfjörð Bjarmason wrote:

Show 21 quoted lines

> [Subject: imap-send: Code correctness flagged by clang]
>
> Clang 1.1 flagged the following issues in imap-send.c, this change
> fixes the warnings by moving some code around:
> 
>     imap-send.c:548:27: warning: data argument not used by format string [-Wformat-extra-args]
>                                cmd->tag, cmd->cmd, cmd->cb.dlen);
>                                                    ^
> 
> Here the sprintf format didn't use the cmd->cb.dlen argument if
> cmd->cb.data was false. Change the code to use a if/else instead of a
> two-level ternary to work it. This code was introduced with imap-send
> itself in f2561fda.
> 
>     imap-send.c:1089:41: warning: conversion specifies type 'unsigned short' but the argument has type 'int' [-Wformat]
>                     snprintf(portstr, sizeof(portstr), "%hu", srvc->port);
>                                                         ~~^   ~~~~~~~~~~
> 
> Here sprintf is being given an int with a %hu format. Cast the
> srvc->port to unsigned short to work it. This code was introduced in
> 94ad2437 to add IPv6 support.

Nitpick: that this was found by clang is probably not the first thing
a person trying to figure out what the patch does needs to know.
Maybe:

	Subject: imap-send: Fix sprintf usage

	When composing a command for the imap server, imap-send
	uses a single nfsnprintf() invocation for brevity
	instead of dealing separately with the case when there
	is a message to be sent and the case when there isn’t.
	The unused argument in the second case, while valid,
	is confusing for static analyzers and human readers.

	v1.6.4-rc0~117 (imap-send: add support for IPv6, 2009-05-25)
	mistakenly used %hu as the format for an int “port”, by
	analogy with existing usage for the unsigned short
	“addr.sin_port”.  Use %d instead.

	Noticed with clang.

Show 16 quoted lines

> +++ b/imap-send.c
> @@ -543,9 +543,14 @@ static struct imap_cmd *v_issue_imap_cmd(struct imap_store *ctx,
>  	while (imap->literal_pending)
>  		get_cmd_result(ctx, NULL);
>  
> -	bufl = nfsnprintf(buf, sizeof(buf), cmd->cb.data ? CAP(LITERALPLUS) ?
> -			   "%d %s{%d+}\r\n" : "%d %s{%d}\r\n" : "%d %s\r\n",
> -			   cmd->tag, cmd->cmd, cmd->cb.dlen);
> +	if (cmd->cb.data) {
> +		bufl = nfsnprintf(buf, sizeof(buf),
> +		                  CAP(LITERALPLUS) ? "%d %s{%d+}\r\n" : "%d %s{%d}\r\n",
> +		                  cmd->tag, cmd->cmd, cmd->cb.dlen);
> +	} else {
> +		bufl = nfsnprintf(buf, sizeof(buf), "%d %s\r\n", cmd->tag, cmd->cmd);
> +	}
> +

Hmm, maybe this would be easier to read:

	if (!cmd->cb.data)
		bufl = nfsnprintf(buf, sizeof(buf), "%d %s\r\n", cmd->tag, cmd->cmd);
	else
		bufl = nfsnprintf(buf, sizeof(buf), "%d %s{%d%s}\r\n",
		                  cmd->tag, cmd->cmd, cmd->cb.dlen,
		                  CAP(LITERALPLUS) ? "+" : "");

i.e., putting the easier case first and avoiding a variable format string.

Show 6 quoted lines

> @@ -1086,7 +1091,7 @@ static struct store *imap_open_store(struct imap_server_conf *srvc)
>  		int gai;
>  		char portstr[6];
>  
> -		snprintf(portstr, sizeof(portstr), "%hu", srvc->port);
> +		snprintf(portstr, sizeof(portstr), "%hu", (unsigned short)srvc->port);

Why not

		snprintf(portstr, sizeof(portstr), "%d", srvc->port);

Thanks for checking the code.