On Sat, Jun 23, 2012 at 10:29:29PM +0900, Osamu Aoki wrote: > This is non-essential but I thought it may be good idea not to make > postinst script robust. As it is written now, any non-zero exit code of > update-catalog will break postinst script.
This is somewhat intentional. If update-catalog breaks, then something really is broken and should fail loudly. > Why not like: > update-catalog --quiet --update-super || > echo "SGML catalog is broken. Read > /usr/share/doc/sgml-base/README.trouble." 1>&2 This would hide failures to write the super catalog and cause very nasty failures as we have seen in the past. > FYI: I did not understand why this script returns > * 1=error for normal errors, > * 0=non-error for goods and > * -1=error (for shell this 255) for --help and --version. > The last type was not expected. I think that update-catalog never had a useful distinction between 1 and -1. For instance it runs into both printing a usage sometimes. (--help causes -1, --doesnotexist causes 1) Note that a "die" results in -1. What kind of failures are "normal errors"? Which failures would you like not to be loud? Helmut -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
