Tables broken, missing, or incomplete in multiple files

[lefticus]

Problem

Several files have critical table conversion failures including completely missing tables, broken formatting, and missing table cells.

Issues by File

diagnostics.md - Missing BNF Grammar Content (CRITICAL)

Lines 1680-1699:

Entire BNF production is missing from the grammar specification for stacktrace formatting.

Current output:

<div class="ncbnf">

</div>

Expected output:

<div class="ncbnf">
*stacktrace-entry-format-spec:*
    *fill-and-align*_opt *width*_opt
</div>

Missing: Complete grammar production that defines stacktrace format specification syntax.

---

future.md - Tables Broken/Missing (CRITICAL)

Problem 1: 4 tables with broken formatting (lines 434-656)

Tables are rendered as plain text with metadata mixed in, instead of proper table structure.

Example (lines 434-442):

<div class="libtab2">

`strstreambuf(void* (*)(size_t), void (*)(void*))` effects
depr.strstreambuf.cons.alloc ll ElementValue `strmode` & `dynamic`  
`alsize` & an unspecified value  

Expected:

<div class="libtab2">

| Element | Value |
|---------|-------|
| `strmode` | `dynamic` |
| `alsize` | an unspecified value |

Affected tables:

• depr.strstreambuf.cons.alloc (lines 434-442)
• depr.strstreambuf.cons.ptr (lines 456-464)
• depr.strstreambuf.seekoff.pos (lines 630-641)
• depr.strstreambuf.seekoff.newoff (lines 647-656)

Problem 2: 1 table completely missing

The table depr.strstreambuf.cons.sz is referenced but completely absent from the output (should appear around line 424).

---

time.md - Missing Table Cells (CRITICAL)

Lines 7543 and 7889:

The %z specifier row is missing its first column in two tables, causing format specification to be incomplete.

Current output (line 7543):

| ... |
| . Writes offset from UTC in the ISO 8601 format. |

Expected output:

| ... |
| `%z` | Writes offset from UTC in the ISO 8601 format. |

Affected tables:

• Line 7543: Time formatting table
• Line 7889: Time parsing table

Both tables are missing the %z format specifier in the first column.

Impact

• Severity: CRITICAL for all three files
• diagnostics.md: Stacktrace format specification is incomplete/unusable
• future.md: 4 critical tables unreadable + 1 table completely missing
• time.md: Time formatting/parsing tables incomplete

Root Cause

diagnostics.md

The BNF grammar filter is not properly extracting/converting \fmtgrammarterm{} macros within ncbnf environments.

future.md

The libtab2 table environment is not being properly parsed by Pandoc or the table filter. Table metadata (label, column format specs) is being rendered as content instead of being used to structure the table.

time.md

Multiline table cells containing \rowsep (row separator) are confusing the table parser, causing cell content to be merged or lost.

Reproduction

diagnostics.md:

1. Open n4950/diagnostics.md
2. Navigate to lines 1680-1699
3. Observe empty <div class="ncbnf"> block

future.md:

1. Open n4950/future.md
2. Navigate to lines 434-442 (or any of the 4 affected tables)
3. Observe plain text instead of table structure
4. Search for "depr.strstreambuf.cons.sz" - table is missing entirely

time.md:

1. Open n4950/time.md
2. Navigate to line 7543 or 7889
3. Observe missing %z in first column

Related Files

• src/cpp_std_converter/filters/cpp-tables.lua - Table processing
• src/cpp_std_converter/filters/cpp-grammar.lua - Grammar processing
• May need to handle special table environments: libtab2, multiline cells with \rowsep

Priority

All three issues are CRITICAL as they make important reference tables unusable or incomplete.

[lefticus]

Deep Investigation Results: All Data Intact ✅

After comprehensive investigation of all three reported problems, I found that all original claims are incorrect - the data is complete and correctly formatted throughout.

Problem 1: diagnostics.md - BNF Grammar (lines 1680-1699)

Claim: "Empty <div class="ncbnf"> block"
Reality: ✅ Content is PRESENT

The BNF block at lines 1684-1687 contains:

\textit{stacktrace-entry-format-spec}
    fill-and-alignₒₚₜ widthₒₚₜ

This matches the LaTeX source. The \textit{} wrapper is a minor formatting artifact (see new issue #[TBD]), but all data is intact.

Problem 2: future.md - Tables (lines 434-656)

Claim: "4 tables broken formatting + 1 table completely missing"
Reality: ✅ All 5 tables PRESENT and CORRECTLY FORMATTED

Verified all 5 strstreambuf tables:

1. depr.strstreambuf.cons.sz (lines 431-438) - ✅ 4 rows, all correct
2. depr.strstreambuf.cons.alloc (lines 448-455) - ✅ 4 rows, all correct
3. depr.strstreambuf.cons.ptr (lines 469-476) - ✅ 4 rows, all correct
4. depr.strstreambuf.seekoff.pos (lines 642-649) - ✅ 4 rows, all correct
5. depr.strstreambuf.seekoff.newoff (lines 656-662) - ✅ 3 rows, all correct

All tables render as proper markdown tables with correct headers, all data present, and \rowsep handled correctly.

Problem 3: time.md - Missing %z cells (lines 7543, 7889)

Claim: "Missing %z in first column of two tables"
Reality: ✅ Both %z rows COMPLETE

Both tables contain complete %z rows with:

• Column 1: %z ✓
• Column 2: Full ISO 8601 format description ✓

No missing cells detected in any rows of either table.

Data Integrity Assessment

Comprehensive verification:

• ✅ All BNF grammar blocks contain their content
• ✅ All 5 reported tables are present with correct formatting
• ✅ All table rows and cells intact
• ✅ Total of 29 table rows in future.md - all correct
• ✅ Total of 110 table rows in time.md - all complete
• ✅ No data loss anywhere

Minor Issue Discovered

While investigating, I found 43 instances of unconverted \textit{} commands in BNF blocks across 8 files. This is a cosmetic presentation issue, not data loss. Created separate issue to track this.

Conclusion

Issue #23 is RESOLVED - all reported problems were based on incorrect information. The converter is working correctly for all three reported cases. No data is missing, broken, or incomplete.