pack_string should never free str on error. This wouldn't be a problem
except the check for `ours` was inverted. Just remove the check for
ours since the true condition is unreachable.
json_vpack_ex also had an error check for s.has_error. This can never
be true unless value is NULL.
Expand pack_unpack testing to cover empty format string, NULL object
value with non-null concatenated key, array containing a non-null object
after a NULL (error) string.
Fixes#444
* Remove errant line-feed from pack_object error message.
* Correct error message in pack_object_inter.
* Create pack_integer / pack_real to get the correct error messages on
failure when packing numeric values.
* Add tests for packing NAN and infinity directly, in an array and as
an object value.
When NULL is received for an optional argument we should not set an
error message as this would block later error messages. If NULL is
received for a non-optional string we should set has_error. Set
has_error for UTF-8 errors to ensure optional strings with UTF-8
errors are not replaced with json_null(). Use 'purpose' argument in
NULL error messages of read_string.
Add error handling and tests for invalid formats where '+', '#', or '%'
is used on an optional string 's?' or 's*'.
Fix NULL string error messages to use 'purpose'.
Refactor skipping of '*' token, this is now handled by read_string and
pack_object_inter. This allows invalid format strings such as 's*#' and
's*+' to produce error messages.
Fixes#437
* Test equality of different length strings.
* Add tab to json_pack whitespace test.
* Test json_sprintf with empty result and invalid UTF.
* Test json_get_alloc_funcs with NULL arguments.
* Test invalid arguments.
* Add test_chaos to test allocation failure code paths.
* Remove redundant json_is_string checks from json_string_equal and
json_string_copy. Both functions are static and can only be called
with a json string.
Fixes to issues found by test_chaos:
* Fix crash on OOM in pack_unpack.c:read_string().
* Unconditionally free string in string_create upon allocation failure.
Update load.c:parse_value() to reflect this. This resolves a leak on
allocation failure for pack_unpack.c:pack_string() and
value.c:json_sprintf().
Although not visible from CodeCoverage these changes significantly
increase branch coverage. Especially in src/value.c where we previously
covered 67.4% of branches and now cover 96.3% of branches.
Users of the "o" format have an expectation that the object reference
will be stolen. Any error causes the collection process to end early.
This patch causes json_pack and related functions to continue scanning
the format and parameters so all references can be stolen to prevent
leaks. This makes no attempt to continue processing if the format
string is broken or missing.
'make check' still passes. Ran test_pack under valgrind and verified
that the leaked reference is fixed. Added a test which uses refcounts
to verify that the reference was correctly stolen after a NULL value
error.
Issue #135
This is because it's really easy to get a name collission if compiling
Jansson as a subproject in a larger CMake project. If one project includes
several subprojects each having their own config.h, this will cause the
wrong file to be loaded.
gcc 4.2.1 warns about a possible incorrect cast to ssize_t when comparing against refcount, which is of type size_t. signed vs unsigned comparison.
Sinc warnings are treated as errors (as in the autoconf project) this would cause a compile error.
- Moved everything to one CMakeLists.txt
- Added support for the json_process test suites (instead of just the API
tests).
- Changed to use the modified json_process version that does away with the
environment variables (originally written by DanielT).
- Had to exclude "test_memory_funcs" on MSVC, since void pointer
arithmetics are not allowed as it is done in secure_malloc and
secure_free.
- Had to add a check for "ssize_t". This is not available on Windows and
maybe on some other platforms (used in test_pack.c)
- Result from running ctest (The failure seems unrelated to CMake, it's
just that the expected result is in a different order):
99% tests passed, 1 tests failed out of 121
Total Test time (real) = 1.31 sec
The following tests FAILED:
24 - valid__complex-array (Failed)
The decimal point '.' is changed to locale's decimal point
before/after JSON conversion to make C standard library's
locale-specific string conversion functions work correctly.
All the tests now call setlocale(LC_ALL, "") on startup to use the
locale set in the environment.
Fixes GH-32.
- Add a new field position to the json_error_t structure. This is the
position in bytes from the beginning of the input.
- Keep track of line, column and input position in the stream level.
Previously, only line was tracked, and it was in the lexer level, so
this info was not available for UTF-8 decoding errors.
- While at it, refactor tests so that no separate "stripped" tests are
required. json_process is now able to strip whitespace from its
input, and the "valid" and "invalid" test suites now use this to
test both non-stripped and stripped input.
Closes GH-9.
Expand the pack/unpack API: json_(un)pack is the simple version with
no error output, json_(un)pack_ex has error output and flags,
json_v(un)pack_ex is a va_list version of json_(un)pack_ex.
Implement unpacking flags:
- JSON_UNPACK_ONLY turns off extra validation, i.e. array and object
unpacking doesn't check that all items are unpacked. This is really
just convenience for not adding '*' after each object and array.
- JSON_VALIDATE_ONLY turns off unpacking, i.e. no varargs are expected
and nothing is unpacked.
* Implement a "scanner" that reads the format string, maintaining state
* Split json_vnpack() to three separate functions for packing objects,
arrays and simple values. This makes it more clear what is being
packed, and the object and array structures become more evident.
* Make the skipping of ignored character simpler, i.e. skip ':' and
',' independent of their context
This patch shaves around 80 lines of code from the original
implementation.
Note that we pass va_list pointers around instead of just va_lists, which
would seem more intuitive. This is necessary since the behaviour of va_lists
passed as function parameters is finicky. Quoth stdarg(3):
If ap is passed to a function that uses va_arg(ap,type) then the value
of ap is undefined after the return of that function.
The pointer-passing strategy is used by Python's Py_BuildValue() for the same
purpose.
