Hi Paul, Paul Eggert <egg...@cs.ucla.edu> writes:
> Also, I'm a bit dubious about all those calls to sigprocmask. Can't > we solve this without having to execute a sigmask-related system call > for each buffer? How about using the method that 'ls' uses instead? > Install a signal handler that merely sets a static variable. Perhaps > the relevant 'ls' code should be Gnulib-ized, so that it can be shared > between 'ls' and 'diff'. is it ok for now to keep these calls to sigprocmask and fix all the other issues you reported (until the ls.c code is Gnulib-ized)? They won't affect any existing use-case anyway as they will be used only when --color is used on a tty. Thanks for the detailed review. Giuseppe
