[OE-core] [PATCH 2/3] pseudo: support multiple search directories in PSEUDO_PASSWD

Peter Seebach peter.seebach at windriver.com
Mon Oct 13 01:30:18 UTC 2014


On Sun, 12 Oct 2014 18:49:54 -0500
"Peter A. Bigot" <pab at pabigot.com> wrote:

> ++		npasswd_paths = build_passwd_paths(NULL);
> ++		if (npasswd_paths) {
> ++			passwd_paths = malloc(npasswd_paths * sizeof(*passwd_paths));
> ++			if (!passwd_paths) {
> ++				pseudo_diag("couldn't allocate space for passwd paths.\n");
> ++				exit(1);
> ++			}
> ++			build_passwd_paths(passwd_paths);

I'm slightly inclined to think that this might be better done with a function
which counts colons and then builds the path list accordingly. But I
basically like the idea of supporting a path list. As I recall, we talked
about this early on, but for some reason I formed the theory that it wasn't
necessary.

I'd point out that strictly speaking, the change from [] to * doesn't matter
in function parameter lists, and I tend to prefer [] when the argument is
logically an array rather than a pointer to a single item, except that this
preference is not something I'm very consistent about.

But this does seem to me to be correct.

I'll have to think about it a bit because, while the "call once to find out
how much storage you need, again to use the storage" idiom is pretty good, I
think its primary benefit over simpler calling patterns is that it can be
used with arbitrary memory allocation patterns or goals. In this case, with
only one caller, we probably don't need it. That said, if I'm still busy this
week (which looks likely) it's not a bad idea to merge this as-is and wait for
1.6.3 for arguable improvements, since this is pretty clearly correct.

-s
-- 
Listen, get this.  Nobody with a good compiler needs to be justified.



More information about the Openembedded-core mailing list