On Wed, Mar 19, 2014 at 08:28:07AM -0600, Eric Blake wrote:
> On 03/19/2014 08:19 AM, Jeff Cody wrote:
>
> >>> + then
> >>> + _timed_wait_for ${h} "${@: -1}"
> >>
> >> You have done shift before this. Aren't ${*} the remaining strings to wait
> >> for ?
> >>
> >
> > I could probably get rid of the 2nd shift, although I would have to
> > adjust the conditional below.
> >
> > I do ${@: -1} because I want the very last whole string to be the item
> > to wait for - this is only needed to accommodate pathnames with spaces
> > inside the QMP string.
>
> ${@: -1} is not portable:
>
> $ bash -c 'set 1 2 3; echo ${@: -1}'
> 3
> $ dash -c 'set 1 2 3; echo ${@: -1}'
> dash: 1: Bad substitution
>
> If you want the last argument, you'll have to do something hideous like:
>
> eval \${$#}
>
> Short of using eval, there is no portable way to get at the last
> positional argument in dash.
>
Yes, and there are likely other bash-isms in some of the shell
scripts in qemu-iotests. Since #!/bin/bash is explicitly specified,
it seems reasonable that bash-isms would be allowed. If it was
#!/bin/sh specified as the interpreter, then I would understand
remaining constrained to POSIX-only.
But I think in your next message you have a nice POSIX compatible
method of doing it with shifts, and it is probably best to default to
POSIX when practical. I'll go ahead and change it to the
'shift $(($# - 1))' method.
