[#8] Revisions and fixes to enable YDBTest imptp tests to run and pass

All YottaDB APIs must be tested using the imptp portion of the YDBTest test framework. This commit includes various revisions and fixes to enable imptp tests to run and pass within YDBTest.See the YDBTest!1299 merge request for those changes.

To allow correct redirection of output in impjob.py instances spawned during YDBTest imptp testing, it was necessary to implement a wrapper for the ydb_stdout_stderr_adjust Simple API function. This function is wrapped by the new adjust_stdout_stderr function, which was implemented in _yottadb.c and added to yottadb.py. It is not straightforward to test directly, so test coverage for this function is accomplished by the imptp tests in YDBTest.

To allow YDBPython to be used and tested with address sanitization, the following changes were made:

• Revised setup.py to automatically compile YDBPython with ASAN when YDB (in $ydb_dist) is compiled with ASAN
• Fixed heap-buffer-overflow detected by ASAN settings by removing erroneous + 1 for null terminator in memcpy() call in _yottadb.c
• Fixed heap-buffer-overflow bug detected by ASAN in ci_wrapper()
  • Added gparam_list struct to _yottadb.h for use in callers of ydb_call_variadic_plist_func()
  • Revised type of arg_values from uintptr_t to gparam_list in callers of ydb_call_variadic_plist_func()
  • Removed malloc() and free() calls, as gparam_list is now stack allocated
  • Updated various type casts and array assignments to match gparam_list member types
• Added missing free() calls in various places

Pipeline updates:

• Revised CentOS jobs to use Rocky instead of CentOS to match other YDB repos
• Broke .test subjob into two subjobs: .test_ubuntu and .test_rocky to allow different ASAN specifications per platform
• Increased test verbosity in pipeline by adding -v to pytest call, allowing overview of tests passing/failing at a glance
• Added redirect of pytest stdout to file and added to job artifacts to facilitate debugging
  • Added error file for pytest output to allow review of pytest's stderr output
• Revised test_wordfreq to output to file instead of stdout and added new output files to artifacts

Test fixes:

• Added try/except block to handle KeyError when environment variables are already unset in test_no_ydb_gbldir(). This resolves failures where the test tries to unset ydb_gbldir but it wasn't set in the environment, causing a KeyError exception. So, an except block catching and ignoring this exception was added to handle that case.
• Decreased sleep time to 0.1 seconds in lock timeout tests to prevent timing-related failures. This was necessary because lock_value is a function that acquires a lock and sleeps for 0.2 seconds before releasing it. Therefore if the affected lines sleep for 0.2 seconds, it is more likely that by the time the _yottadb.lock_incr() call in each is done, the initial process has released the lock. In that case no timeout error is raised and the test fails. Reducing the timeout to 0.1 seconds (half of the time that process sleeps with the lock held) eliminates this issue by almost guaranteeing process is still sleeping while holding the lock.
  • This prevents the test from waiting so long that the lock is released before an attempt is made to acquire it.
• Revised test_unsigned_int_length_bytes_overflow to use 2**32 (i.e. UINT32_MAX + 1) instead of (2**32) + 1 (i.e. UINT32_MAX + 2) to test the off-by-one case
  • Corrected type size check in anystr_to_buffer() by replacing INT32_MAX with UINT32_MAX
  • Updated comments in test_unsigned_int_length_bytes_overflow and anystr_to_buffer()

README.md updates:

• Updated README with package instructions
• Added cleanup instructions to README
• Added ASAN setup to README
• Added comments explaining lock_timeout tests sleep duration
• Fixed tab/spaces issue and restructured instructions for clarity
• Fixed list numbering in various places that were out of order and made one item lists unordered instead of numbered lists
• Changed secondary headers to use ## (h2) instead of # (h1) for clearer organization and consistency with HTML best practice

Miscellaneous changes:

• Added tmp* to .gitignore to ignore auto-generated tmp.mupip file and test subdirectories used for test database creation
• Replaced various return values of FALSE with !YDB_OK to make return value success/failure format consistent with YDB API calls, which return YDB_OK on success
• Renamed YDB_INSTALL_DIR to YDB_DIST
• Added missing function information to docstring for tp()
• Revisions to fix failure due to erroneous counting of '^' as part of global variable name during buffer population in _yottadb.c
  • Refactored POPULATE_NEW_BUFFER and cast_pyssize_t_to_unsigned_int into existing anystr_to_buffer
  • Removed a bunch of dead code and unused variables, simplified calls to PyArg_ParseTupleAndKeywords by reducing argument count
  • Added new YDBPY_ERR_ARG_NOT_BYTES_LIKE for use in generic TypeErrors during argument parsing
• Fixed bug in functions calling the POPULATE_SUBS_USED_AND_SUBSARRAY() macro wherein the value of return_null was not checked across macro calls, allowing failures in prior macros to go undetected. This is now fixed by the following changes:
  • Refactored POPULATE_SUBS_USED_AND_SUBSARRAY() into populate_subs_used_and_subsarray() function
  • Replaced above macro with new function and relevant error handling throughout _yottadb.c
  • Refactored affected functions to add explicit error checking, cleanup, and remove implicit fall-throughs via return_null variable
• Corrected SIG-11 in set() due to unallocated default value, by adding default value and conditional free() in the case where the default is NOT used
• Added default Python parameter initialization to some functions
• Added INT32_MAX overflow check to anystr_to_buffer
• Reduced duplication by moving buffer and subsarray population and cleanup into new INVOKE_ANYSTR_TO_BUFFER and INVOKE_POPULATE_SUBS_USED_AND_SUBSARRAY_AND_CLEANUP_VARNAME macros
• Fixed compiler warnings on freeing of string literals by converting these literals into heap allocated variables (the compiler did not respect/understand the logic that prevented this scenario and issued this warning anyway)
• Split YDBTimeoutError into YDBLockTimeoutError and YDBTpTimeoutError to distinguish timeouts issued by lock() and tp()
• Revised conftest.py to store both stdout and stderr output in execute() function for use in debugging
• Removed newlines before returns in _yottadb.c