[pacman-dev] [arch-dev-public] [signoff] pacman 3.5.0-1

[Dave Reisner]

On Wed, Mar 16, 2011 at 03:29:24PM -0500, Dan McGee wrote:
> On Wed, Mar 16, 2011 at 3:26 PM, Aaron Griffin <aaronmgriffin at gmail.com> wrote:
> > On Wed, Mar 16, 2011 at 3:24 PM, Dan McGee <dpmcgee at gmail.com> wrote:
> >> On Wed, Mar 16, 2011 at 3:18 PM, Gaetan Bisson <bisson at archlinux.org> wrote:
> >>> [2011-03-16 15:10:08 -0500] Dan McGee:
> >>>> 2. This is a terrible bug report. :/ I need output when you run this
> >>>> by hand, maybe with --debug, etc. Help me help you by actually
> >>>> providing details.
> >>>
> >>> I thought it was reproducible well enough. There is no debug output:
> >>>
> >>>        yes | pacman -S --debug
> >>>
> >>> prints nothing and just eats up RAM. I will try to investigate more and
> >>> let you know if I find anything.
> >>
> >> Damn! This is a huge regression disguised as a "feature" that I never
> >> quite felt comfortable with...
> >> http://projects.archlinux.org/pacman.git/commit/?id=4fb3cfc48f626f84329c78351d6a8710255a35e3
> >
> > You can do it "GNU style" and only read from stdin if the package list is "-"
> 
> OMG, Aaron is here!
> 
> That sounds like the way to go, or we are going to cause regressions
> to anyone that expected prior behavior. Dave- does that make sense,
> only read from stdin if an '-' arg is provided?
> 
> -Dan

A wild Griffin appears! This sounds like it would be a good compromise,
but getopt won't allow this. It would need to be parsed as a target,
which also means that it needs to be the _only_ target. Therefore, if...

!isatty(fileno(stdin)) && alpm_list_count(targets) == 1 &&
  alpm_list_getdata(targets) == '-'

...then we read stdin. It's a bit magical, but I'm happy to implement
and update the doc if this is an acceptable fix.

dave