krb5 commit: Comment why TRACE() is not used directly

[Benjamin Kaduk]

https://github.com/krb5/krb5/commit/014bc344f8f6307c97b0b23326600cbc8ae1a236
commit 014bc344f8f6307c97b0b23326600cbc8ae1a236
Author: Ben Kaduk <kaduk at mit.edu>
Date:   Sat Jul 23 14:59:05 2016 -0500

    Comment why TRACE() is not used directly
    
    Instead, a new TRACE_* macro should be defined for each trace point, and
    that macro used from the library code.

 src/include/k5-trace.h |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/src/include/k5-trace.h b/src/include/k5-trace.h
index 045f102..c75e264 100644
--- a/src/include/k5-trace.h
+++ b/src/include/k5-trace.h
@@ -28,6 +28,11 @@
  * This header contains trace macro definitions, which map trace points within
  * the code to krb5int_trace() calls with descriptive text strings.
  *
+ * A new trace macro must be defined in this file for each new location to
+ * be traced; the TRACE() macro should never be used directly.  This keeps
+ * the tracing logic centralized in one place, to facilitate integration with
+ * alternate tracing backends such as DTrace.
+ *
  * Trace logging is intended to aid power users in diagnosing configuration
  * problems by showing what's going on behind the scenes of complex operations.
  * Although trace logging is sometimes useful to developers, it is not intended