-------- Original Message --------
From: Otto Moerbeek <o...@drijf.net>
To: bytevolc...@safe-mail.net
Cc: tech@openbsd.org
Subject: Re: [PATCH] Option for mount_tmpfs to populate the volume
aftercreation.
Date: Thu, 18 Sep 2014 10:54:34 +0200
> On Thu, Sep 18, 2014 at 08:40:44AM +0100, bytevolc...@safe-mail.net wrote:
>
> > This patch adds an option "-t template" to mount_tmpfs, which
> > populates the new tmpfs volume with a directory
> > immediately after creation.
> >
> > Man page update included for explanation.
> >
> > Much of the code was grafted from newfs
> > which implements this for mount_mfs ("-P" option).
> >
> > Suggestions, fixes, criticism, etc. welcome.
>
> - Please do not mix the argument parsing changes with the new
> functionality changes. Post separate diffs for that,
Thanks for the info. Will fix set of diffs.
> although I do not
> see the point of removing mount_tmpfs_parseargs().
My rationale for this is as follows: in order for the new code to
work, I need to store the user-provided template path somewhere.
I could add an extra parameter to mount_tmpfs_parseargs() as one of the
following:
1. char ** - risk optarg may change after the function has
returned, assuming wanted to call getopt() after calling
mount_tmpfs_parseargs() later on when they edit the code further
2. char* - point to yet another buffer allocated just to store the template
path. Much better than above.
Rather than going through one of the above, I thought a neater solution would
be how newfs handles this, so I integrated mount_tmpfs_parseargs() into its
caller. The code for tmpfs_parseargs() itself isn't removed; just relocated.
That being said, if having this as a separate function will make the overall
code easier to maintain, I'll work around it. It does not seem too difficult.
> - Why use a different flag compared to mount_mfs?
Mostly a semantics thing IMO. I just thought it made more sense to use "-t" as
in "template", since most file attributes are copied over, as opposed to just
"let's populate the directory with a bunch of files."
As I was updating the man page I was thinking about whether to use "-t" or
"-P", but then I figured "let's see what the devs think first."
The next diff will end up using "-P" instead then based on your input.
> - What's pathnames.h doing here?
I thought it would be a good idea to include it in the diff in the same
directory as I wasn't sure if this would compile on another person's machine;
it compiles and works fine on mine though.
I wasn't sure how include directories (#include "...") worked in the source
tree so this was a just-in-case measure. If not needed, I am somewhat relieved.
> -Otto
- Adam.
