Re: [PATCH] Add ability to specify environment extension to run_command

From: Junio C Hamano <junkio@cox.net>
Date: May 22, 2007, 06:33 UTC
Message-ID: <7v646l9xkn.fsf@assigned-by-dhcp.cox.net>
In-Reply-To: <20070522060302.GH5412@admingilde.org>

Martin Waitz <tali@admingilde.org> writes:

Show 8 quoted lines

> On Mon, May 21, 2007 at 04:02:42PM -0700, Junio C Hamano wrote:
>> I had a feeling that some callers needed to be able to unsetenv
>> some.  How would this patch help them, or are they outside of
>> the scope?
>
> At first I had the same objection but the putenv documentation
> told me that at least in glibc you can unsetenv by providing
> the variable name without a "=".

I recall SysV putenv() does not remove "ENVNAME" without '=', and http://www.opengroup.org/onlinepubs/000095399/functions/putenv.html seems to say that as well.

> But perhaps we should check for other systems?

Probably.

At least we have setenv/unsetenv calls already (with emulation where they aren't available), we could do something like this:

	struct child_process {
        	...
                const struct {
                        const char *name;
                        const char *value; /* NULL to unsetenv */
                } *env;
		...
	};

and in start_command():

	if (cmd->env) {
		int i;
                for (i = 0; cmd->env[i].name; i++) {
			if (cmd->env[i].value)
				setenv(cmd->env[i].name, cmd->env[i].value, 1);
			else
				unsetenv(cmd->env[i].name);
                }
        }