On Sat, Jun 21, 2008 at 4:56 PM, James Carlson <james.d.carlson at sun.com> 
wrote:
> Mike Gerdts writes:
>> I would like to get some eyes on my proposed fix for 6717067.
>>
>> http://cr.opensolaris.org/~mgerdts/6717067/
>
> 217: Too bad X11 has 'bogus share' disease ... I think I would have
>     used /usr/X11/man here instead (as that's really the proper path
>     for a man page), but it probably doesn't matter.  (By "bogus
>     share," I'm referring to the fact that /usr/share is a special
>     mount location defined in filesystem(5) ... putting your own
>     directory under there is reasonable, but putting "share" into
>     some far-flung path makes no sense at all.  Nobody's going to
>     share a bunch of random directories.  In short, "/usr/share/foo/"
>     is right, and "/usr/foo/share" is wrong.)

I agree, mostly.  I chose /usr/X11/share/man because that is a
directory, whereas /usr/X11/man is a symlink.  They both get you to
the same place, but /usr/X11/share/man seemed to be chosen as the
"right" place.  To be honest, I am just thrilled that the "right"
place for X executables is not /usr/bin/X11/*.

> 3186: the logic appears to be backwards here.  The original code used
>      "!=0" for the stat() check -- it was returning NULL when the
>      directory being stat'd didn't exist.  This new code is returning
>      NULL when the directory _does_ exist, and non-NULL when it
>      doesn't.
>
>      The code should be simply:
>
>        if (stat(mand, &sb) == 0 && S_ISDIR(sb.st_mode))
>                return (mand);
>
>        free(mand);
>        return (NULL);

Oops.  Fixed.

> Other than that, it looks like this file doesn't (or shouldn't) pass
> C-style, as it has non-compliant function definitions.  Oh, well.
> Don't bother fixing that if you don't feel up to it.

"cstyle -pP man.c" doesn't complain.  A read of
http://opensolaris.org/os/community/documentation/getting_started_docs/cstyle.ms.pdf
suggests that this may be what you are referring to:

static void
freev(char **v) {

should be:

static void
freev(char **v)
{

I've fixed those. Is there something else I should be looking for?

FWIW, there are also a bunch of continuation line warnings (cstyle
-c).  Since Roland has indicated that he will be suggesting a project
to tackle some rather significant work with the man subsystem, I will
leave this for now.

I have updated the webrev at http://cr.opensolaris.org/~mgerdts/6717067/

Thanks for the review!
Mike

-- 
Mike Gerdts
http://mgerdts.blogspot.com/

Reply via email to