Gisle Vanem wrote:
No. He added a 'reset_color_context (true);' inside
'signal_handler()'.
Ah, sorry, I was looking at the original edition of the patch.
I just now looked at <http://bugs.gnu.org/20062#101> and still see some
problems. Mainly, it doesn't handle SIGTSTP properly. With SIGTSTP an
application is supposed to reset the terminal and then resignal itself with
SIGSTOP before really stopping. When restarting the application then needs to
restore the terminal to the current state, if the current state is not the
default. This is nearly impossible to do correctly with the proposed design. I
suggest looking at how GNU ls does it: the signal handler merely sets a flag,
and the application checks occasionally at safe places whether a signal has arrived.
A nit: when installing signals via 'signal', don't install a handler if 'signal'
says the signal is being ignored.
Also, copyright papers need to be filed for diffutils before a nontrivial patch
like this can be installed.
