On Mon, 2015-12-07 at 15:10 +0000, Ian Jackson wrote:
> Ian Campbell writes ("Re: [OSSTEST PATCH 04/11] mg-debug-fail: New
> utility script for debugging"):
> > On Fri, 2015-12-04 at 19:35 +0000, Ian Jackson wrote:
> > > diff --git a/mg-debug-fail b/mg-debug-fail
> > > new file mode 100755
> > > index 0000000..64fa235
> > > --- /dev/null
> > > +++ b/mg-debug-fail
> > > @@ -0,0 +1,13 @@
> > > +#!/bin/sh
> > > +#
> > > +# This script can be provided anywhere an executable or command name
> > > is
> > > +# wanted. It prints its arguments, and its stdin, to its stderr,
> > > and
> > > +# then exits nonzero.
> > > +#
> > > +# When using this it may be useful to provide </dev/null as a
> > > +# redirection for the whole program under test. Otherwise things
> > > +# can mysteriously hang.
> >
> > "egrep . - /dev/null" is too noisy, it adds a "(standard input):"
> > prefix
> > which I don't think you want. But "egrep -h . - /dev/null" seems to
> > remedy
> > this without the possibility of these mysterious hangs.
>
> This is confusing to me. The problem is that mg-debug-fail does not
> know whether to try to read and print its stdin. Sometimes its stdin
> will be the stdin for some utility that it is standing in for, and
> should be dumped. Whereas sometimes its stdin is the user's tty,
> which should be left alone.
>
> This difficulty is fixed in the next patch. None of your suggestions
> seem to address it. What problem do you think they are solving ?
I think I got off on a bit of a tangent, and hadn't seen the next patch
yet.
> > Also, what does egrep give us here over just cat?
>
> I meant to write egrep ''
> and its advantage is that it sanitises missing trailing newline.
>
> But maybe we want to preserve missing trailing newline.
I don't think so, I just didn't know about that behaviour of egrep, it
sounds useful.
>
> An alternative to replace the egrep would be
> sed 's/^/mg-debug-fail-stdin: /' >&2
> which preserves missing trailing newline but annotates the output.
>
> Ian.
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
