Suggested merge order for rc2

[fdcastel]

The open PRs for an rc2 and the order in which they merge cleanly, smallest and most isolated first. This list used to live in a comment on #302; it is kept up to date here as PRs open and merge.

1. Fix SQLPrepare discarding SQL_ATTR_ROWS_FETCHED_PTR and SQL_ATTR_ROW_STATUS_PTR (#301) #303 — two lines in OdbcDesc.cpp (setDefaultImplDesc), plus a regression test. Fixes SQLPrepare discards SQL_ATTR_ROWS_FETCHED_PTR and SQL_ATTR_ROW_STATUS_PTR set before it, so block cursors never learn how many rows a fetch returned #301.
2. Return the settable pointer attributes and SQL_ATTR_CURSOR_SCROLLABLE from SQLGetStmtAttr #304 — eight case lines in sqlGetStmtAttr (OdbcStatement.cpp), plus a round-trip test.
3. tests: rowset fetches on every fetch path and binding layout #305 — tests only, stacked on Fix SQLPrepare discarding SQL_ATTR_ROWS_FETCHED_PTR and SQL_ATTR_ROW_STATUS_PTR (#301) #303; one commit once Fix SQLPrepare discarding SQL_ATTR_ROWS_FETCHED_PTR and SQL_ATTR_ROW_STATUS_PTR (#301) #303 is in.
4. Apply the bind offset to column-wise rowsets instead of taking the row-wise path (#306) #315 — column-wise rowsets with a bind offset, stacked on tests: rowset fetches on every fetch path and binding layout #305; two commits in fetchData, its static-cursor twin and returnDataFromExtendedFetch. Fixes Column-wise binding with SQL_ATTR_ROW_BIND_OFFSET_PTR writes every row of a rowset to the same address #306.
5. Keep SQL_ATTR_KEYSET_SIZE apart from the rowset size; HY007 for an unprepared IRD (#307) #314 — SQL_ATTR_KEYSET_SIZE kept apart from the rowset size, HY007 for an unprepared IRD; three commits on master. Fixes SQL_ATTR_KEYSET_SIZE overwrites the rowset size; SQLGetDescField on an unprepared IRD answers HY091 instead of HY007 #307.
6. Implementation descriptors and SQLColAttribute report the columns and parameters they describe (#316) #317 — implementation descriptors and SQLColAttribute report what they describe, stacked on Keep SQL_ATTR_KEYSET_SIZE apart from the rowset size; HY007 for an unprepared IRD (#307) #314; ten commits, tests first, in OdbcDesc.cpp, OdbcDesc.h, sqlColAttribute in OdbcStatement.cpp and the two SQLColAttribute entry points in Main.cpp and MainUnicode.cpp. Fixes Implementation descriptors: records empty until the first bind or execute, raw type fields, 32-bit length writes; SQLColAttribute type fields #316.
7. Fix numeric parameters silently sent as NULL after a character-typed bind (#300) #302 — one commit: describe an INPUT parameter's precision from the prepare-time snapshot. Fixes SQL_BIGINT parameter: once a NULL has been bound with a character C type (SQL_C_DEFAULT or SQL_C_CHAR), later SQL_C_SBIGINT rebinds on that parameter write NULL #300.
8. Fix column-wise parameter arrays reading element 0 for fixed-length C types (#299) #308 — parameter arrays: three commits, tests first; 22 driver lines in bindInputOutputParam, sqlExecute / sqlExecDirect and getConciseSize. Fixes Parameter arrays: fixed-length C types use BufferLength as the element stride, so every row receives element 0 (silent data corruption) #299.
9. Honour SQL_ATTR_PARAM_OPERATION_PTR in parameter-array execution #310 — SQL_ATTR_PARAM_OPERATION_PTR, stacked on Fix column-wise parameter arrays reading element 0 for fixed-length C types (#299) #308; two commits, one function.
10. Return the cumulative row count after a parameter-array execute #311 — cumulative SQLRowCount after an array execute, stacked on Honour SQL_ATTR_PARAM_OPERATION_PTR in parameter-array execution #310; two commits.
11. Restore the bind-offset pointer when a parameter-array execute throws (#309) #313 — bind-offset pointer restored when an array execute throws, stacked on Return the cumulative row count after a parameter-array execute #311; two commits, same function. Fixes Statement handle crashes on reuse after a server error inside a parameter-array execute #309.
12. Fix SQL_C_GUID parameter binding sending corrupted data on the wire (#295) #296 — SQL_C_GUID parameter binding, all review notes addressed. Fixes SQL_C_GUID parameter binding sends corrupted data on the wire (Phase 8 / T5-5) #295.
13. Fix silent data loss binding numeric to VARCHAR via stored procedure #292 — numeric to VARCHAR through stored procedures, three small commits.

All thirteen are green on the full matrix.

Driver-code hunks do not overlap between independent PRs, with one adjacent-line exception: #314 and #311 each add a member after rowNumberParamArray in OdbcStatement.h, so whichever lands second needs a one-line rebase. Otherwise: #308 shares OdbcDesc.cpp with #303 and OdbcStatement.cpp with #304, in different functions; #310, #311 and #313 change executeStatementParamArray in sequence, which nothing else touches; #315 changes the fetch path, which nothing else touches; #314 adds a case fifty lines from the block #304 adds; #314 and #317 change sqlGetDescField in sequence; #317's sqlColAttribute hunk is in a function nothing else touches. Shared files on the test side: #292 and #302 both add helpers to test_param_conversions.cpp; #303, #304 and #305 each add a line to tests/CMakeLists.txt; #303, #305 and #315 extend test_block_cursor.cpp in sequence; #308, #310, #311 and #313 extend test_array_binding.cpp in sequence; #314 and #317 extend test_descriptor.cpp in sequence, and #314 alone touches test_scrollable_cursor.cpp. Whichever lands later gets a rebase from me the same day.

Every open issue reported this month has a PR above.

[fdcastel]

An unofficial build of all thirteen is now in my fork's Releases section, for anyone who wants to try them together before they land here:

https://github.com/fdcastel/firebird-odbc-driver/releases/tag/v3.5.1-rc2

It is master (35fae3e) with the PRs above merged on top, built by the same workflow that produces the official artifacts: MSI and ZIP for Windows x64, x86 and ARM64, tar.gz for Linux x64 and ARM64. The full matrix is green on the merged branch.