Russ Cox writes:
>
> first, a fair bit of the code assumes that unsigned and
> signed characters are the same; of course it doesn't
> help that gcc can't tell them apart.
Some of that had already been fixed, I've checked in fixes for the rest.
> second, the concepts of mknod and rdev are foreign to
> plan 9. you might want to make an ST_RDEV() macro
> to handle the latter; i'm happy if you don't handle the former:
> i can always define a null one.
Most of that code is only compiled if PRESERVE_PERMISSIONS_SUPPORT is
defined, which it isn't in the current development version (since the
code is notoriously buggy). My inclination is to add #ifdef's around
the remaing code as well, but I'd like a little education, too: How does
one create device files in plan 9 without mknod()? How does one
determine whether two device files are identical without st_rdev? You
just commented out the code that doesn't compile, I'd rather have code
that compiles and does the right thing!
> third, lib/savecvd.c includes <fcntl.h> but forgets to include <sys/types.h> first.
> you will probably want more #ifdef's than i used.
I just put it inside the #ifdef for HAVE_FCNTL so it's only included
when <fcntl.h> is included.
-Larry Jones
I think your train of thought is a runaway. -- Calvin's Mom
