From: Stefan Sperling <stsp@stsp.name>
Subject: Re: inttypes.h?
To: "Todd C. Miller" <Todd.Miller@sudo.ws>
Cc: Christian Weisgerber <naddy@mips.inka.de>, gameoftrees@openbsd.org
Date: Sat, 26 Sep 2020 20:49:03 +0200

On Sat, Sep 26, 2020 at 09:51:38AM -0600, Todd C. Miller wrote:
> On Sat, 26 Sep 2020 17:41:58 +0200, Christian Weisgerber wrote:
> 
> > The diff below
> >
> > * ... casts time_t and off_t values to long long, so they are
> >   guaranteed to match the %lld format specifier.  On OpenBSD that
> >   happens to be the case, on FreeBSD it's a mess that varies by
> >   architecture.  I don't know if that's acceptable, but I see similar
> >   casts in ssh.
> >   (Style? Second change in got_object_commit_create() below.)
> 
> I think this is about the best you can do.  One alternative is to
> cast to intmax_t instead of long long, though I don't think it
> really matters.
> 
> > * ... switches the number in parse.y from int64_t to long long,
> >   since it is assigned to from strtonum(), passed to parseport()'s
> >   long long parameter, and printed with %lld anyway.
> 
> That makes sense to me and argues for using long long above.

I agree. This isn't a burden at all. We'll just need to try to keep
these explicit casts in mind when handling time_t and off_t. Since the
patch is consistently adding such casts in all relevant places now,
this should be easy to keep in mind going forward.