8.2 bug when moving files around

Joe Pruett joey at q7.com
Thu Jun 3 01:52:18 UTC 1999


i've stumbled across a bug in 8.2 in the following scenario.

i'm secondary for domains from site x.  so i use lines like the following
in my config:

zone "foo.com" { type slave; file "x/foo.com"; masters { 192.168.1.1; }; };

the name of the site changes to y, so i change the config to reflect that:

zone "foo.com" { type slave; file "y/foo.com"; masters { 192.168.1.1; }; };

and then i rename the x subdir to y (since i have lots of domains).  i
then do a reload and things seem fine.  eventually one of the domains changes,
and a named-xfer gets spit out like this:

named-xfer -z foo.com -f y/foo.com -i x/foo.com.ixfr.tmp -s 1 -C 1 -P 53 -T foo.com.120 192.168.1.1 axfr

which generates a syslog message like this:

named-xfer[17894]: rename y/foo.com.0ZHiJb to x/foo.com.ixfr.tmp: No such file or directory

it looks like end_zone is calling validate_zone with a pre-existing
structure, so validate_zone doesn't regenerate the ixfr_tmp field.
which is probably because the whole method for reloading doesn't throw
away old data before re-reading the config file.  it looks like a few
other fields will get treated the same way in validate_zone.  i don't
see a trivial fix, so i'll throw it out here.

is this one fixed in the mysterious 8.2.1?



More information about the bind-users mailing list