On FreeBSD 13.0-CURRENT its compiler clang-11.0.0 complains:
index.c:388:14: error: comparison of integers of different signs: 'int' and 'unsigned int' [-Werror,-Wsign-compare]
for (i=0; i<len+1; i++) {
~^~~~~~
* Separate -D (date only) and -T (date/time) format options in mdb-export and mdb-json
* New public mdb_set_shortdate_fmt() function in libmdb
* New private(ish) mdb_col_is_shortdate() function
I'm calling it "shortdate" in order to preserve the existing API.
See https://github.com/mdbtools/mdbtools/issues/12
This should fix long-standing complaints about the default bind size
without causing undue memory inflation in existing applications.
Could make this adjustable on the command line later.
Supersedes:
https://github.com/mdbtools/mdbtools/pull/137
Quickstart (requires Clang 6 or later):
$ export LIB_FUZZING_ENGINE=/path/to/fuzzing/library.a
$ ./configure --enable-fuzz-testing
$ make
$ cd src/fuzz
$ make fuzz_mdb
$ ./fuzz_mdb
Also add a new `mdb_open_buffer function` to facilitate in-memory
fuzz-testing. This requires fmemopen, which may not be present on all
systems. The internal API has been reworked to use file streams instead
of file descriptors. This allows reading from memory and reading from
files using a consistent API.
There are more modern tools for memory debugging, get rid of DMALLOC
crap in the source code.
I've left one reference in backend.c to prevent a merge conflict but
this can be removed later.
* Replace backend_is_init (formerly is_init) with relationships_table
* Remove deprecated functions that used static variables
* Move static "bound" variable to MdbHandle as relationships_values
Attempt to make the backend handling logic thread-safe. This removes the
last MDB_CONSTRUCTOR. Also get rid of some JAVA junk and make the
remaining static variables in backend.c constant. Finally remove some
obsolete fields from MdbFile.
I'm not sure if this is a complete solution - some of the global state
has been moved to thread-local storage. So passing an ODBC handle across
threads may have unexpected results. But at least it's not global state.
Each ODBC handle now has its own iconv_t object.
See #23
