Greg Hudson [Thu, 5 Feb 2009 19:59:09 +0000 (19:59 +0000)]
In krb5_rc_io_store, check the return value of krb5int_buf_len as well
as krb5int_buf_data. The length can't be negative if the data is
non-NULL, but Coverity doesn't know that.
Greg Hudson [Thu, 5 Feb 2009 19:50:41 +0000 (19:50 +0000)]
In recvauth_common, initialize ap_option. It can't be used
uninitialized, but you can only deduce that by examining the
relationships between ap_option, problem, and outbuf.
Greg Hudson [Thu, 5 Feb 2009 19:44:35 +0000 (19:44 +0000)]
Change krb5_rc_resolve_type (not a public API) to allocate the rcache
structure. Make output parameter values of krb5_rc_resolve_type and
krb5_rc_default well-defined in case of errors.
Greg Hudson [Thu, 5 Feb 2009 18:34:57 +0000 (18:34 +0000)]
In krb5_ktfileint_find_slot, check for an error return from ftell.
(Such an error will never happen in any reasonable stdio
implementation but it's more correct to check.)
Greg Hudson [Thu, 5 Feb 2009 18:26:47 +0000 (18:26 +0000)]
Coverity was nervous that hst_realm.c's domain_heuristic() wasn't
checking for a NULL return from strchr. The code was safe because a
previous call to strchr on the same argments was checked, but make
Coverity less nervous by storing the result of that previous call and
reusing it. Also make the function conform better to our standards.
Ken Raeburn [Thu, 5 Feb 2009 02:16:22 +0000 (02:16 +0000)]
declare replacement [v]asprintf functions
If HAVE_VASPRINTF is not defined, make sure krb5int_{,v}asprintf
functions always get declared, applying the preprocessor conditional
test only to the GCC format attribute. If HAVE_VASPRINTF is defined,
don't declare them at all.
This fixes a bunch of function-not-declared warnings under Sun cc.
Ken Raeburn [Thu, 5 Feb 2009 01:49:21 +0000 (01:49 +0000)]
int/ptr bug in gssapi code
Fix a pointer argument passed where an integer is needed.
Update Sun compiler options to make that an error. (The options we're
currently using make it an error for assignment but not for argument passing.)
Greg Hudson [Wed, 4 Feb 2009 20:32:05 +0000 (20:32 +0000)]
Remove xfread/xfwrite macros. Casting the first argument to char * is
unnecessary (fread's first argument is void *, which does not require
a cast) and confuses Coverity's UNINIT checker; casting the third
argument to unsigned is not necessary for our current set of warnings.
Ken Raeburn [Mon, 2 Feb 2009 21:13:42 +0000 (21:13 +0000)]
Get rid of casts of free() argument to char*, except where it's
casting away const (so as to make this change warning-neutral), and in
unicode source (which we may want to keep in sync with another
source), and krb5_xfree macro (to be handled separately).
Ken Raeburn [Mon, 2 Feb 2009 20:37:41 +0000 (20:37 +0000)]
Revise last change to better resemble the original test, keeping the
min-lifetime test code collected together. Change policy to have a
minimum password lifetime of 10s instead of 30s, and reduce the test
delays accordingly.
Ken Raeburn [Mon, 2 Feb 2009 19:29:52 +0000 (19:29 +0000)]
speed up kpasswd tests
Reorder some tests and tune delays, so that we don't need to run for
much more than twice the min-password-life interval when testing that
functionality. (This could be made faster if we can assume that
init_db will always have been run immediately before the tests start.)
In my tests, this cuts something like 11 seconds off the run time (now
down to about 65 seconds).
Ken Raeburn [Mon, 2 Feb 2009 18:42:06 +0000 (18:42 +0000)]
address lib/kadm5 test suite slowness
In mod-principal tests for clearing the principal's policy, instead of
just testing to see if the wrong string is output and timing out
looking for it, check also for the new expected value. Cuts test
suite run time by about two minutes for each pass (client vs server).
Ken Raeburn [Sat, 31 Jan 2009 04:39:34 +0000 (04:39 +0000)]
use t_inetd with a ready message and avoid waiting a lot in non-root tests
Change t_inetd to print a ready message when it has started listening
on the indicated port number.
Look for this message in sample.exp rather than waiting an arbitrary
(and usually excessive) 2s each time for the inetd-mode tests. Use
run_once to perform the standalone-mode test only once per test suite
invocation.
Change rsh and rcp tests to start the servers via t_inetd and avoid
excessive waiting at startup.
In some of my tests, this reduces the tests/dejagnu tests from taking
over 6 minutes to taking around 2 minutes.
(This does mean the server process will no longer have started up
before we launch the client, so it may be slower to respond, but it'll
still be faster than the 2s delay we used before even trying to
connect.)
We can probably eliminate the -D option code from krshd.c now.
The tests run as root (rlogin, telnet) still need updating.
Will Fiveash [Fri, 30 Jan 2009 23:55:14 +0000 (23:55 +0000)]
Master Key Migration Project
Commit for the Master Key Migration Project.
http://k5wiki.kerberos.org/wiki/Projects/Master_Key_Migration
This commit provides the ability to add a new master key (with an
enctype differing from the current master key) to the master key
principal and stash file and then migrate the encryption of existing
principals long term keys to use the new master key. In addition
deletion of master keys is provided.
Ken Raeburn [Wed, 28 Jan 2009 00:04:15 +0000 (00:04 +0000)]
Start to phase out krb5_xfree macro, which just casts its argument to
char* and calls free.
Replace most uses, outside of the LDAP KDB plugin, which doesn't build
on my test system of the moment because of version dependencies. Add
one explicit cast to make the change warning-neutral (under gcc 4.0.1
on Mac OS X 10.5.6).
Ezra Peisach [Sun, 25 Jan 2009 16:44:02 +0000 (16:44 +0000)]
Use a struct in_addr to insure alignment of address - instead of
random alignment on the stack. Solaris 2.10 has issues if the address
is not aligned. The rest of the code in the tree uses a struct
in_addr or mallocs the address - which will be sufficiently aligned.
Ken Raeburn [Sat, 24 Jan 2009 01:27:16 +0000 (01:27 +0000)]
Change 16/32/64-bit big-/little-endian/native unaligned load/store
routines to take void pointers, so they can operate on both plain and
unsigned char buffers, or other types.
Remove some now-unneeded casts.
Ken Raeburn [Fri, 23 Jan 2009 18:04:08 +0000 (18:04 +0000)]
hash db2 code breaks if st_blksize > 64k
The hash db code assumes in places that the block size is no larger
than 64K. There's a range check in the case where you don't have a
file but provide initialization info. The btree code will cap the
block size used at 64K.
Apparently Sun's ZFS can report back a block size of 128K, causing the
db2 tests to fail.
Add such a cap to the hash db creation code.
Note that our default configuration is to use the btree code when
creating a new database, so it's unlikely that this will cause
real-world problems unless someone went out of their way to specify
use of the hash format.
Ken Raeburn [Fri, 23 Jan 2009 00:46:46 +0000 (00:46 +0000)]
Include cksumtypes.h from aead.h to get struct krb5_cksumtypes (moved
in rev 21753).
Protect cksumtypes.h from multiple inclusions.
Update dependencies.
projects / thirdparty / krb5.git / log
