Issue 7208: HALT mode (lwlog-rtf) Source: (Mr. Dominick Paniscotti, ) Nature: Uncategorized Issue Severity: Summary: The desired behavior of the log service when the log full action type is set to WRAP needs to be explicitly defined. This comment is based upon experience with the SCA 2.2 JTeL JTAP tests. Open interpretation of this item led to failures characterized by the following questions: a. Is a log with a log full action type of WRAP ever considered full? The JTeL JTAP tests expect this, but there is a semantic issue here – writes to a “full log” will be successful. Further, if a log in WRAP mode can become full, then… b. When is it considered full? Since log records use variable amounts of storage, a log nearing capacity may have a handful of bytes available, yet not have enough space to store even the smallest of records. c. When should the log service detect that the log is full? JTAP expects a log to be set full as soon as storage is exhausted, i.e., at the completion of a write that exactly fills the log. While this does not seem to be what the specification suggests, it illustrates an example of an area left up to interpretation. Declaring a log full only in HALT mode and only when a write is attempted would avoid these questions. Resolution: Revised Text: Actions taken: March 25, 2004: received issue Discussion: End of Annotations:===== Subject: RE: November 2003 Version 1.0, formal/03-11-03 Issue Date: Thu, 25 Mar 2004 14:13:08 -0500 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: November 2003 Version 1.0, formal/03-11-03 Issue Thread-Index: AcQSnKdzIFEZ2XQjRGK3V+E0HpNkqwAACucAAAAFF8AAAASj4AAAByrQ From: "Paniscotti, Dominick (US SSA)" To: , Cc: "Fay, Thomas J. (US SSA)" X-OriginalArrivalTime: 25 Mar 2004 19:13:08.0936 (UTC) FILETIME=[3576CC80:01C4129D] The desired behavior of the log service when the log full action type is set to WRAP needs to be explicitly defined. This comment is based upon experience with the SCA 2.2 JTeL JTAP tests. Open interpretation of this item led to failures characterized by the following questions: a. Is a log with a log full action type of WRAP ever considered full? The JTeL JTAP tests expect this, but there is a semantic issue here . writes to a .full log. will be successful. Further, if a log in WRAP mode can become full, then. b. When is it considered full? Since log records use variable amounts of storage, a log nearing capacity may have a handful of bytes available, yet not have enough space to store even the smallest of records. c. When should the log service detect that the log is full? JTAP expects a log to be set full as soon as storage is exhausted, i.e., at the completion of a write that exactly fills the log. While this does not seem to be what the specification suggests, it illustrates an example of an area left up to interpretation. Declaring a log full only in HALT mode and only when a write is attempted would avoid these questions.