[reporter="jvs", created="Wed, 20 Feb 2008 01:18:24 -0500 (GMT-05:00)"]
I am putting in some special casing in NativeTrace so that the loggers we create for Segments and ExecStreams don't leak (when disabled) due to their per-object logger names.
New leaks could happen if additional tracing is added to Fennel with per-object logger names, so this issue is here to remind me that a more general scheme is needed. And maybe at the same time come up with a way to avoid leaks even when these are enabled?
Maybe the per-object naming should go as a prefix on each message (rather than as part of the logger name)?
[reporter="jvs", created="Wed, 20 Feb 2008 01:18:24 -0500 (GMT-05:00)"] I am putting in some special casing in NativeTrace so that the loggers we create for Segments and ExecStreams don't leak (when disabled) due to their per-object logger names.
New leaks could happen if additional tracing is added to Fennel with per-object logger names, so this issue is here to remind me that a more general scheme is needed. And maybe at the same time come up with a way to avoid leaks even when these are enabled?
Maybe the per-object naming should go as a prefix on each message (rather than as part of the logger name)?