It broke a sequence of:
key1 \
value1 \
value2 \
value3 \
key2 \
value4 \
etc.
Handle both the netgroup rule and the amd.home rule using a .USEBEFORE
macro rule, since they contain the same code.
XXX: Needs new make(1)
because neither "ypxfr" nor the YP bits of "libc" can handle IPv6. If the
hostname is used there will be an automatic fallback to IPv4 which makes
server pushed work again in mixed IPv4 and IPv6 environments. This fixes
bin/11755 by myself.
define _LIBC so that the weak aliases those files normally define if
built in libc get set properly. Otherwise, this doesn't link statically.
Static linking tested on NetBSD/i386 by sommerfeld, correct runtime
operation not verified (but should be no different than before).
access checking (via libwrap) like mountd(8) does.
Also, garbage-collect a bunch of stuff that is normally created by
rpcgen(1) that we'll never use in this program.
There might be serveral more waiting to be pushed to.
This is only 1/2 of the fix. ypxfer(8) also needs a change to
avoid a false error, but this allows pushes to continue to work,
at least.
Problem noted by aflury@zembu.com and agordon@zembu.com.
hosts.{byname,byaddr} is still an IPv4 only mapping.
obeys solaris8 practice. see comments in ypinit/Makefile.yp for details.
TODO: interop test with solaris8 (any takers?)
through sort before makedbm.
This drops Erik Rungi's <rungus@openface.ca> passwd.byname generation for
3000 entries from 25-30 seconds down to 0.75 seconds.
the same (datum is { char *, int }, DBT is { void *, size_t /* u_int */ }),
copy info to appropriate temporary variables before calling underlying
db(3) operations.
- ensure hostname from gethostname() is nul-terminated in all cases
- minor KNF
- use MAXHOSTNAMELEN over various other values/defines
- be safe will buffers that hold hostnames
i've fixed a lot of bugs in it):
* implement yp_close_map(struct opt_map *), to close a specific map,
using the guts of yp_close_last()
* simplify yp_close_last() to use yp_close_map()
* ypdb_open_db(): add the st_dev, st_ino, and st_mtime of each map to
the cached info, and check if changes if the db is cached; if so, close
the map and reopen it fresh. also uncache entries if the domain or
file disappear
if (test1)
if (test2)
error()
else {
...
}
this happened when i changed test2 from a void statement to actually
checking its return value.
the effect of this? a YP_MASTER_KEY value wasn't being added to the
generated databases, which was Not Good.
- use fgetln() instead of fgets()
- store info in internal buffer
- fix \\ support
- count line numbers internally, so \\ lines don't mess up count
* ypdb_store():
- ensure that the length of key or val doesn't exceed YPMAXRECORD
* makedbm, mknetid, revnetgroup, stdethers, stdhosts:
- improve error handling
- take advantage of rewritten read_line(), and cleanup line parsing
- don't print trailing ` ' for key/val pairs with an empty val
* Makefile.yp:
- fix up building of ypservers (from Chuck Cranor)
* ypinit.sh:
- remove leading spaces in variable assignment (from Chuck Cranor)
- rename OPTDB to OPTIMIZE_DB
- if OPTIMIZE_DB, in ypdb_open(), close a database if the open failed
with ENFILE or EMFILE [from openbsd].
- because of the above, don't bother checking first if any FDs are
available (with open("/", RDONLY) - it's a superfluous check.
- ensure that *map_info is set if a database was a cached open -
without this fix OPTIMIZE_DB didn't work too well...
- enable OPTIMIZE_DB by default - should speed up lookups
secure map support:
- add ypdb_secure(), which returns non-zero if the map has the YP_SECURE
private key set (which makedbm(8) -s adds) [from openbsd]
- modify various functions to check if a map is secure, and if so,
refuse to serve information to clients not requesting from an
IPPORT_RESERVED port [from openbsd]
- ensure that domainnames and map names are valid (don't contain
"/"'s, aren't too long, etc) [idea from openbsd, but using our
_yp_invalid_domain() et al, which does more checking.]
misc:
- add rcsid's
- implement _yp_invalid_map() (complements _yp_invalid_domain() in
libc - eventually _yp_invalid_map() may move to libc...)
- use %#x instead of 0x%x in debug log messages
follow our formatting convetions.
- add support for aliases, amd.home, master.passwd and netgroup maps
[initially from openbsd, reworked as above]
- if INSECURE != "yes" then the passwd.by* maps will not contain the
encrypted version of the passwd - privileged clients will request
info from the master.passwd.by* maps.
FreeBSD originally implemented this, and & OpenBSD supports it too.
[from openbsd].
By default, passwd maps are INSECURE, but in a *BSD-only environment,
this could be disabled to allow for a [slightly] higher degree of
security from non-root users...
if the initial open fails with EFTYPE. if a different database
format is found to be more efficient, just add the code to open it
prior to that of the other formats; this maintains backward
compatibility with older databases whilst providing new functionality...
- add rcsid's
