Tom Yu [Wed, 21 Jun 2006 21:08:31 +0000 (21:08 +0000)]
cease export of krb5_gss_*
* src/lib/gssapi/libgssapi_krb5.exports: Cease export of krb5
mechanism entry points which were once equivalent to GSS-API entry
points. These were never public interfaces, and calling them
directly now can cause problems. Also, cease export of some
other internal symbols.
ticket: new
target_version: 1.5
tags: pullup
component: krb5-libs
Ken Raeburn [Wed, 21 Jun 2006 00:28:21 +0000 (00:28 +0000)]
(GET_HOST_BY_NAME, GET_HOST_BY_ADDR) [GETHOSTBYNAME_RETURNS_INT]: Test
the output hostent pointer as an additional way of checking for errors.
Original patch from Nalin Dahyabhai, needed some tweaking to fit in
current sources, especially as updated in ticket 3911.
Ken Raeburn [Tue, 20 Jun 2006 23:59:11 +0000 (23:59 +0000)]
getaddrinfo code uses vars outside of storage duration
The calls to gethostbyname_r and friends were wrapped in macros that declared
automatic variables for the auxiliary storage needed, but the pointers returned
by the functions would be used outside of that scope, when the storage would no
longer be valid.
Changed the macro interfaces to define new types for the auxiliary storage, and
add a new argument to the macros referring to that storage. Changed call sites
accordingly.
Ken Raeburn [Tue, 20 Jun 2006 23:27:01 +0000 (23:27 +0000)]
(kadmin_cpw): Don't crash if principal name is not given. Based on a
patch from S Rahul.
(extended_com_err_fn): If error code is zero, don't fetch an error message.
Russ Allbery [Tue, 20 Jun 2006 21:45:16 +0000 (21:45 +0000)]
In verify_krb_v4_tgt, use an unsigned KRB4_32 to store an address rather
than an unsigned long to avoid memcpying past the end of a structure on
64-bit platforms.
Russ Allbery [Mon, 19 Jun 2006 23:37:09 +0000 (23:37 +0000)]
Pass in the correct value for the first argument of select (one larger
than the largest file number in the select set) rather than some multiple
of sizeof some struct. The latter is large enough accidentally work, but
breaks on AIX.
Map IPv4-mapped IPv6 addresses back to IPv4 in krshd for the purposes of
connecting back to the remote system on AIX, since on AIX getnameinfo
returns such addresses but connect won't accept them.
Jeffrey Altman [Mon, 19 Jun 2006 17:33:36 +0000 (17:33 +0000)]
Export krb5_gss_register_acceptor_identity in KFW
krb5_gss_register_acceptor_identity is a gss krb5 extension that is
part of the public ABI. It does not have a gss_krb5_* name due to
historical reasons. Instead there is a gss_krb5_register_acceptor_identity
macro that uses this export.
Ken Raeburn [Fri, 16 Jun 2006 00:26:49 +0000 (00:26 +0000)]
Remove -lkrb5support, which should be for internal use only, from the displayed
list of libraries. This means the output is not suitable for linking against
static libraries built from the tree, but prevents us from committing to
keeping libkrb5support.1 in all future Mac shared-library builds (in addition
to the libkrb5support.0 we're stuck with for non-Mac UNIXy shared-library
builds).
Ken Raeburn [Fri, 16 Jun 2006 00:10:08 +0000 (00:10 +0000)]
ftpd's getline conflicts with current glibc headers
Reported by Ubaidul Khan on kerberos@mit this date.
GNU libc includes a function getline with an incompatible prototype.
It hasn't caused problems for MIT on RHEL 4 (glibc 2.3.4), but
apparently FC5 has a newer set of headers in which it does.
* ftpcmd.y (getline): Renamed to ftpd_getline. Declarations and callers
changed.
* ftpd.c (getline): Remove duplicate declaration.
Ken Raeburn [Thu, 15 Jun 2006 22:01:57 +0000 (22:01 +0000)]
(kadmin_startup): Properly skip "\@" and "\/" sequences when
processing principal name, rather than going into an infinite loop.
Suggested by Mark Phalan at Sun.
Ken Raeburn [Wed, 14 Jun 2006 23:44:48 +0000 (23:44 +0000)]
Some simple transformations:
Use ANSI C function definitions instead of old K&R style.
Rename local variable "res" to "clnt_res".
Cast clnt_call arguments to xdrproc_t and caddr_t.
This brings the code more in line with what current (RHEL) rpcgen would do.
Russ Allbery [Wed, 14 Jun 2006 20:31:52 +0000 (20:31 +0000)]
When NULL is passed into krb_get_in_pw_tkt, we only want to prompt once
for the password rather than passing NULL along to each string to key
function causing each to prompt independently. Modify krb_get_in_pw_tkt
to call des_read_pw_string directly and then pass the resulting password
into each string to key function as needed. Add a prototype of
des_read_pw_string to krb4int.h since it's an exported function of
libdes425 but isn't prototyped in des.h.
Russ Allbery [Tue, 13 Jun 2006 16:03:12 +0000 (16:03 +0000)]
telnet help should telnet to a host named help, not display a usage
message. Only support ?, not help, to request help with the open
command. The same change was made in netkit's telnet.
Russ Allbery [Tue, 13 Jun 2006 15:18:04 +0000 (15:18 +0000)]
POSIX allows getpwnam_r and getpwuid_r to return 0 (success) even if the
username or UID could not be found, in which case OUT will be set to
NULL. Elsewhere, code assumes that if k5_getpwnam_r or k5_getpwuid_r
returns 0, OUT is non-NULL. Check whether OUT is NULL and adjust the
return value accordingly in the k5_getpw{nam,uid}_r wrappers.
Russ Allbery [Mon, 12 Jun 2006 19:59:35 +0000 (19:59 +0000)]
Always include sys/ioctl.h in the telnet and telnetd code, even if
sys/filio.h is available. GNU/kFreeBSD has sys/filio.h but still
requires sys/ioctl.h, and sys/ioctl.h is included unconditionally
elsewhere in the code. Patch from Petr Salinger.
Russ Allbery [Mon, 12 Jun 2006 19:47:52 +0000 (19:47 +0000)]
Never return -I/usr/include from --cflags. It's already in the default
compiler search path, causes gcc warnings, and can cause obscure failures
with gcc by including system headers before gcc's internal headers.
Russ Allbery [Mon, 12 Jun 2006 18:58:01 +0000 (18:58 +0000)]
Always initialize the output token in gss_init_sec_context as required
by RFC 2744 section 5.19. The krb5 code did this but the generic code
didn't, causing a double-free in OpenSSH.
Russ Allbery [Mon, 12 Jun 2006 18:35:03 +0000 (18:35 +0000)]
If krb5_gss_canonicalize_name is given GSS_C_NULL_OID as the mechanism,
assume krb5 using similar logic as other library functions rather than
failing. Patch provided by Daniel Kahn Gillmor.
Russ Allbery [Mon, 12 Jun 2006 18:19:26 +0000 (18:19 +0000)]
Document the authorization logic used in the absence of .k5login files.
Remove all remaining references to .rhosts authentication; the code to
implement this was already removed.
Ken Raeburn [Sat, 10 Jun 2006 01:05:40 +0000 (01:05 +0000)]
Since it appears we don't actually set the conf_tgs_ktypes field except in this
initial allocation of zero elements, it can be deleted, along with
conf_tgs_ktypes_count and the associated code...
Ken Raeburn [Thu, 8 Jun 2006 20:23:17 +0000 (20:23 +0000)]
Fixed by having krb5_fcc_interpret call krb5_set_error_message to incorporate
the lower-level error message into the message displayed by kinit, which has
already been changed to use krb5_get_error_message. For example, "kinit(v5):
Credentials cache I/O operation failed (No space left on device) when
initializing cache".
Ken Raeburn [Fri, 2 Jun 2006 23:21:12 +0000 (23:21 +0000)]
No longer ignore configure, autom4te.cache, config.status, config.log
in directories that no longer have configure.in files. (And prune
blank lines previously accidentally added to the svn:ignore property
while we're editing them anyways.)
Sam Hartman [Fri, 2 Jun 2006 21:14:35 +0000 (21:14 +0000)]
Patch from Alejandro R. Sedeno and Jeffrey Hutzelman to allow krb4 to
read 32-bit and 64-bit ticket files on 32-bit and 64-bit systems.
Previously the ticket file format depended on the ABI. Significant
backward compatibility is maintained; the patch works by writing
alignment records that are valid (but meaningless) ticket file entries
but that allow systems to get realigned. As a consequence an old
library will see additional meaningless ticket file entries when it
reads a ticket file produced by the new code. These entries are
harmless and will be ignored.
Ken Raeburn [Wed, 31 May 2006 00:09:31 +0000 (00:09 +0000)]
Merge about 1/3 of the remaining configure scripts into the top level. This
still leaves out appl and tests, and static library and plugin directories.
Ken Raeburn [Wed, 24 May 2006 16:18:53 +0000 (16:18 +0000)]
Drop major version number from service-locator function table structure.
Add a typedef name for the structure. Change multiple-inclusion macro name.
Add a comment indicating desired symbol name.
Ken Raeburn [Tue, 23 May 2006 00:37:46 +0000 (00:37 +0000)]
Mention the python plugin stuff in configure.in, but inside an "if false"
block, so reconf will build a configure script and it'll be easy to enable
for testing.
Ken Raeburn [Tue, 23 May 2006 00:03:06 +0000 (00:03 +0000)]
install headers into include/krb5
Create include/krb5 directory, and put krb5.h and (k5-)locate.h there in the
build tree. Stub krb5.h in main include directory just includes krb5/krb5.h.
Update dependencies, and add dependencies in a couple Makefiles that didn't
have them.
Ken Raeburn [Thu, 18 May 2006 04:22:15 +0000 (04:22 +0000)]
* configure.in: Set and substitute KSU_LIBS, SETENVOBJ, DO_TCL. Generate
makefiles for util/et, util/profile, lib/kdb, clients and those clients
subdirectories built on UNIX, rather than running configure there.
Deleted configure.in scripts for those directories, and changed
Makefile.in definitions of thisconfigdir and mydir.
projects / thirdparty / krb5.git / log
