Commit graph

159 commits

Author SHA1 Message Date
Dag-Erling Smørgrav
f26cbe997d There is no reason not to distribute our ctype-like macros and strl*()
functions.
2015-01-21 23:34:52 +00:00
Dag-Erling Smørgrav
342ac3fd78 Note that rand_bytes() is just a placeholder for the sake of liboath. 2015-01-08 12:24:19 +00:00
Dag-Erling Smørgrav
a33f58cad7 Fix equality predicate. 2015-01-07 19:23:38 +00:00
Dag-Erling Smørgrav
d0fb2359f1 Implement mpi_{add,sub}() in terms of mpi_{add,sub}_abs(). 2015-01-07 18:59:44 +00:00
Dag-Erling Smørgrav
cddb6c0e02 Rewrite the zero shortcut tests to avoid dropping a const qualifier. 2014-12-30 20:03:58 +00:00
Dag-Erling Smørgrav
3904e809b2 Add two test cases for shortcut logic used when one operand is zero and
the other is also the target.
2014-12-30 19:57:07 +00:00
Dag-Erling Smørgrav
241f524238 Fix a bogus t_malloc_fail in one of the mpi_add_abs() tests.
Add several subtraction test cases and modify some others to better
exercise the carry logic, which is far more complex for subtraction than
for addition.

Add test cases where the target is the same MPI as one or both operands.
2014-12-30 19:48:57 +00:00
Dag-Erling Smørgrav
7a90ca7b42 Since X may point to the same MPI as either G or L (or even both), we
can't set the carry flag based on the contents of G or L after having
modified X.  Instead, compute and stash the new carry value first.
2014-12-30 19:14:59 +00:00
Dag-Erling Smørgrav
8d0e24dee7 Remove the "trailing garbage" version of the mpi_cmp() test. It would
be better to systematically pad MPIs used in all tests with garbage.

Greatly expand the number of mpi_cmp() tests to cover all possible
combinations of same or different sign, same or different msb and same
or different magnitude.
2014-12-30 17:00:56 +00:00
Dag-Erling Smørgrav
f18b915b67 Fix the "x = a + b" allocation failure test. Since a and b were
equal, mpi_sub_abs() would return early and the test would never
exercise the mpi_grow() call which is intended to fail.
2014-12-30 16:46:44 +00:00
Dag-Erling Smørgrav
0d483f20ab Fix msb calculation. We may have to backtrack quite a bit, for
instance when subtracting two nearly equal large numbers.
2014-12-30 16:45:18 +00:00
Dag-Erling Smørgrav
2027ce33e7 Add tests for the off-by-one bug we just fixed in mpi_cmp_abs(). 2014-12-30 16:38:47 +00:00
Dag-Erling Smørgrav
1c024928f9 Fix off-by-one bug in mpi_cmp() which could cause it to read past the
end of the buffer.
2014-12-30 16:35:13 +00:00
Dag-Erling Smørgrav
43b870dc83 WIP: implement subtraction correctly. The actual result is now correct
most of the time, but the MSB calculation is not.
2014-12-30 11:23:30 +00:00
Dag-Erling Smørgrav
e21452d0c7 Add a few tests for subtraction. They're not particularly well thought
out, because currently subtraction does not work at all, but they work
well enough in the sense that they fail with the current implementation.
2014-12-30 11:19:02 +00:00
Dag-Erling Smørgrav
7524a14386 comment nit 2014-12-30 11:15:28 +00:00
Dag-Erling Smørgrav
c5b962387c Add mpi_eq_abs() and mpi_eq() predicates. 2014-12-30 11:12:11 +00:00
Dag-Erling Smørgrav
b6368af502 Re-add $(RT_LIBS) where needed. 2014-12-29 23:56:22 +00:00
Dag-Erling Smørgrav
03867777c6 Add numerous tests, mostly for allocation failures and edge conditions.
Add a t_compare_mpi() predicate which checks that two MPIs have the same
value (but not necessarily the exact same representation).
Clean up and improve existing tests.
2014-12-29 23:54:02 +00:00
Dag-Erling Smørgrav
7b6b493509 Declare t_malloc_fail_after. 2014-12-29 23:49:50 +00:00
Dag-Erling Smørgrav
3ec87c7a8c Introduce t_malloc_fail_after: if non-zero, it serves as a countdown to
setting t_malloc_fail.  Also, enforce t_malloc_fail in realloc(), but not
before checking that the argument was valid.
2014-12-29 23:49:29 +00:00
Dag-Erling Smørgrav
bc8e9d67ba In both mpi_add_abs() and mpi_sub_abs(), make sure that the target is
untouched if the operation fails.  Also ensure that A = A +/- A works.
2014-12-29 23:44:34 +00:00
Dag-Erling Smørgrav
98c6349d4a In mpi_zero(), save a memset() if the target is already zero. 2014-12-29 23:22:54 +00:00
Dag-Erling Smørgrav
1412dac680 Shifting any number by zero places is a no-op, but so is shifting zero by
any number of places.
2014-12-29 23:20:26 +00:00
Dag-Erling Smørgrav
8228f19d5a In mpi_copy(), any failure should leave the target untouched. 2014-12-29 23:18:57 +00:00
Dag-Erling Smørgrav
b2ec312d9f Link each test with only the libraries it needs. This avoids rebuilding
every single test whenever any of the libraries changes.
2014-12-29 22:17:02 +00:00
Dag-Erling Smørgrav
889412ec40 Accept a -jN argument which we pass on to gmake.
Delete all coverage files, not just the root one.
Exit if the tests fail so we can inspect the result.
2014-12-29 17:04:30 +00:00
Dag-Erling Smørgrav
85defb9929 Fix a cast that discards a const qualifier. 2014-12-29 15:23:52 +00:00
Dag-Erling Smørgrav
82f5c5cf77 In mpi_load(), use be32dec() when possible.
In mpi_set(), it is impossible for the value being loaded to exceed
the minimum size of an mpi; thus, there is no need for mpi_grow().
2014-12-29 15:08:09 +00:00
Dag-Erling Smørgrav
7a09b0638f I had forgotten that t_add_test() is actually a printf()-like variadic
function, which obviates the need for asprintf() in the string comparison
and equality tests.
2014-12-29 14:24:08 +00:00
Dag-Erling Smørgrav
3333bd5709 Add test cases for the static inline functions in <cryb/endian.h>. 2014-12-29 14:20:28 +00:00
Dag-Erling Smørgrav
3e3d1f8084 Ignore test.cov 2014-12-29 13:38:32 +00:00
Dag-Erling Smørgrav
0dae3b2417 Test string_equal() using the same test cases as for string_compare().
Improve test case descriptions.
2014-12-29 13:36:06 +00:00
Dag-Erling Smørgrav
e20e76a3f7 Recognize "on" and "off". 2014-12-29 13:25:44 +00:00
Dag-Erling Smørgrav
82e32027af Move the decision of whether to check for memory leaks from the malloc()
code to the driver and allow an environment variable to override the
default (on unless coverage is enabled).
2014-12-29 13:24:44 +00:00
Dag-Erling Smørgrav
56eccb99fe Add two utility functions which check if a string contains something that
can be interpreted as either true or false.  Use this instead of a basic
non-NULL test for the CRYB_PERFTEST check.
2014-12-29 13:22:53 +00:00
Dag-Erling Smørgrav
559f6c24f7 Fix off-by-one error in TAP output: the test case number was being
incremented (converted from zero-based to one-based) twice.
2014-12-29 13:03:08 +00:00
Dag-Erling Smørgrav
eb10ce6d05 If the test case changes desc, it *must* point to an allocated buffer
which the driver will free after printing the result.
2014-12-29 12:56:35 +00:00
Dag-Erling Smørgrav
79967f2844 Fix string_equal() for strings of unequal length. 2014-12-29 12:46:44 +00:00
Dag-Erling Smørgrav
2ad45cfccf Add a string comparison function and a string equality predicate.
Write test cases for the former.
2014-12-29 12:41:39 +00:00
Dag-Erling Smørgrav
1a76a0cbb0 Get latest mkpkgng script from OpenPAM. 2014-12-20 02:58:11 +00:00
Dag-Erling Smørgrav
f82bbc1400 Catch up with OpenPAM's latest OATH changes:
Add an oath_mode(3) function which translates from mode names to numbers.

 Consistently use UINT_MAX, not -1, to indicate an invalid response.

 Change the meaning of the window parameter to always indicate the number
 of codes to check *in addition* to the current code.  Note that for TOTP,
 the window goes in both directions; a window of 1 means to check the
 current code plus the previous and next.
2014-12-20 01:45:01 +00:00
Dag-Erling Smørgrav
4b808bc4df Fill in. 2014-12-20 01:38:57 +00:00
Dag-Erling Smørgrav
eadeeb486c Skip memory leak checks if coverage analysis is enabled. 2014-12-20 01:30:01 +00:00
Dag-Erling Smørgrav
75daba7501 Define CRYB_COVERAGE to 1 when building with coverage analysis. 2014-12-20 01:09:58 +00:00
Dag-Erling Smørgrav
13bad71153 Make coverage.sh executable. 2014-12-20 00:52:37 +00:00
Dag-Erling Smørgrav
24ae6c9434 Differentiate between the Cryb libraries and the Cryb.to project. 2014-12-20 00:47:21 +00:00
Dag-Erling Smørgrav
a98753b6ef Fix various cast-qual bugs. 2014-12-20 00:44:24 +00:00
Dag-Erling Smørgrav
dfa89d80d8 Remove bogus props. 2014-12-20 00:40:07 +00:00
Dag-Erling Smørgrav
6ed802ddf3 Consistently use #if HAVE_FOO instead of #ifdef HAVE_FOO. The reason is
that AC_CHECK_DECLS([foo]), unlike AC_CHECK_FUNCS([foo]), will always
define HAVE_FOO, so #ifdef HAVE_FOO will always be true even if it is 0.
This commit finally fixes the [bl]e{32,64}{enc,dec} issue on Linux.
2014-12-20 00:38:54 +00:00