FBTableStatisticsManager.getTableStatistics: Invalid clumplet buffer structure: buffer end before end of clumplet - clumplet too long [SQLState:HY000, ISC error code:335544382]

[EPluribusUnum]

When a database has a lot of tables, the FBTableStatisticsManager.getTableStatistics call fails.
The getTableCount() returns 1367, thus bufferSize() returns 65641, which is bigger than the 64K suppored by the FB API.

Can you reduce getDatabaseInfo() call buffer size requirement by getting the stats one-by-one ( getInfoItems() )?

[EPluribusUnum]

My unoptimized workaround

try {
	int bufferSize = bufferSize(getTableCount());
	if (bufferSize > 65535) {
		Map<String, TableStatistics> ret = new HashMap<>();
		Map<String, TableStatisticsBuilder> builders = new HashMap<>();
		for (byte ii : getInfoItems()) {
			long size = 1 + 1/*8*/ * (3 + 6 * (long) (getTableCount() + tableSlack));
			if (size <= 0) {
				bufferSize = Integer.MAX_VALUE;
			}
			bufferSize = (int) Math.min(size, Integer.MAX_VALUE);
			Map<String, TableStatistics> resp = db.getDatabaseInfo(new byte[] { ii }, bufferSize, tableStatisticsProcessor);
			for (Entry<String, TableStatistics> e : resp.entrySet()) {
				TableStatisticsBuilder builder = builders.computeIfAbsent(e.getKey(), k -> TableStatistics.builder(k));
				long value = 0;
				switch (ii) {
					case isc_info_read_seq_count:
						value = e.getValue().readSeqCount();
						break;
					case isc_info_read_idx_count:
						value = e.getValue().readIdxCount();
						break;
					case isc_info_insert_count:
						value = e.getValue().insertCount();
						break;
					case isc_info_update_count:
						value = e.getValue().updateCount();
						break;
					case isc_info_delete_count:
						value = e.getValue().deleteCount();
						break;
					case isc_info_backout_count:
						value = e.getValue().backoutCount();
						break;
					case isc_info_purge_count:
						value = e.getValue().purgeCount();
						break;
					case isc_info_expunge_count:
						value = e.getValue().expungeCount();
						break;
				}
				builder.addStatistic(ii, value);
			}
		}
		for (Entry<String, TableStatisticsBuilder> e : builders.entrySet()) {
			ret.put(e.getKey(), e.getValue().toTableStatistics());
		}
		return ret;
	}
	return db.getDatabaseInfo(getInfoItems(), bufferSize, tableStatisticsProcessor);
}
catch (InfoTruncatedException e) {
	/* Occurrence of truncation should be rare. It could occur if all tables have all statistics items, and
		new tables are added after the last updateMapping() call or statistics were previously requested by
		a different instance and this instance was created after tables have been dropped.
		Here, tableSlack is incremented to account for tables removed, while updateTableMapping() is called
		to account for new tables. */
	tableSlack++;
	updateTableMapping();
	lastTruncation = e;
}

[mrotteveel]

The limit of buffer size is not 65535 (at least, not in recent Firebird versions). Unless I've missed something, that limit only applies to individual information items (i.e. more than 10,000 tables should be possible). Can you post the full exception stacktrace?