On Mon 30 Jul 2012 at 03:40PM, Shawn Walker wrote:
> >I'll just write it to stderr instead-- it seems more sane than
> >inventing yet another new exception type. The other option would
> >be to print nothing, which would also be fine with me.
>
> The silent approach may be best; it should be pretty obvious if the
> terminal type is screwed up since the progress tracking will be
> different. And if it isn't obvious, then the caller needs to check
> that.
Ok, so I have a new version of this fix posted. Here are the
changes:
- Now silent on a TERM it doesn't understand; so the
core fix is now very trivial.
- Turns out the previous tests sometimes interacted badly
with the api suite-- mostly due to the way libcurses
initializes itself; once you've called setupterm() once,
subsequent calls in the same process have no effect, which
makes testing various edge cases impossible. So I
reimplemented the tests in the CLI suite; this necessitated
adding a rough variant of Popen and of cmdline_run which
allows one to launch tests in a pty. I also migrated a
pkglint test which did essentially the same thing to use the
new API.
It's the usual "3 lines to fix the bug, 100 lines to test the fix."
https://cr.opensolaris.org/action/browse/pkg/dp/setupterm/
-dp
--
Daniel Price, Solaris Kernel Engineering
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
