Correction for passing zero-length SQL to trace

[Noremos]

Postfix for #8738

[dyemanov]

I'm OK with the correction for master, but I'm not sure we can backport std::string_view into v5. Will you provide a separate patch for v5?

[Noremos]

I'm OK with the correction for master, but I'm not sure we can backport std::string_view into v5. Will you provide a separate patch for v5?

Done: #8748

[asfernandes]

IMO this if wih attribution of m_string_len is very confusing.

[Noremos]

Is this option suitable (more code, but cleaner execution):

diff --git a/src/jrd/trace/TraceDSQLHelpers.h b/src/jrd/trace/TraceDSQLHelpers.h
index c8c2a230bf..a47d5aacca 100644
--- a/src/jrd/trace/TraceDSQLHelpers.h
+++ b/src/jrd/trace/TraceDSQLHelpers.h
@@ -55,11 +55,17 @@ public:
 
 		m_start_clock = fb_utils::query_performance_counter();
 
-		static constexpr std::string_view empty_string = "<empty statement>";
-		if (m_string == nullptr || (m_string_len == 0 && (m_string_len = fb_strlen(m_string)) == 0))
+		if (m_string == nullptr)
 		{
-			m_string = empty_string.data();
-			m_string_len = empty_string.length();
+			traceEmptyStatement();
+		}
+		else
+		{
+			if (m_string_len == 0)
+				m_string_len = fb_strlen(m_string);
+
+			if (m_string_len == 0)
+				traceEmptyStatement();
 		}
 	}
 
@@ -105,6 +111,13 @@ public:
 	}
 
 private:
+	void traceEmptyStatement()
+	{
+		static constexpr std::string_view empty_string = "<empty statement>";
+		m_string = empty_string.data();
+		m_string_len = empty_string.length();
+	}
+
 	bool m_need_trace;
 	Attachment* m_attachment;
 	jrd_tra* const m_transaction;

[dyemanov]

What about this:

if (m_string == nullptr)
	traceEmptyStatement();
else if (m_string_len == 0)
{
	m_string_len = fb_strlen(m_string);
	if (m_string_len == 0)
		traceEmptyStatement();
}

[Noremos]

Looks good. I applied these changes.