What is the Shore Release?

Release 1.1.1 of the Shore software represents a bug-fix and maintenance release relative to the previous release. The release is targeted for sites that will use Shore as a system for writing applications using the Shore Data-definition Language and for those who want to write their own servers using the Shore Storage Manager. This document describes the release and provides pointers to other documents explaining how to install and use the software.

If you have questions, comments or bug reports, please mail them to

    shore_support@cs.wisc.edu

For information on other mailing lists, see the Shore WWW home page: http://www.cs.wisc.edu/shore/ .  

Requirements

The Version 1.1.1 release of Shore is supported for Solaris 2.5 on SPARCstations and Intel X86 (Pentium) workstations and Linux. Shore has been ported to other platforms, but there are no supported ports at this time. We hope to make available, on an "as is" basis, ports contributed by the user community.

You will need a current version of GNU C/C++ compiler gcc (with its include files) and the GNU make utility. We do not support other compilers or versions of make . The current release builds successfully with versions 2.7.2.2 and 2.7.2.3 of the compiler. The pre-compiled (binary) release was built with gcc 2.7.2.3 and libg++ 2.7.2. As newer versions of the compiler are released, we will update the Shore software.

Complete details about required software and hardware are available in the Requirements section of the Installation manual.

 

Copyright and Disclaimer

The Shore software is distributed with the following copyright notice and disclaimer, and is subject to its terms:

Copyright © 1994-7
Computer Sciences Department, University of Wisconsin-Madison.
All Rights Reserved.

Permission to use, copy, modify and distribute this software and its documentation is hereby granted, provided that both the copyright notice and this permission notice appear in all copies of the software, derivative works or modified versions, and any portions thereof, and that both notices appear in supporting documentation.

THE COMPUTER SCIENCES DEPARTMENT OF THE UNIVERSITY OF WISCONSIN - MADISON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" CONDITION. THE DEPARTMENT DISCLAIMS ANY LIABILITY OF ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.

The Shore Project Group requests users of this software to return any improvements or extensions that they make to the Shore Project Group, Computer Sciences Department, University of Wisconsin, 1210 West Dayton Street, Madison, WI 53706-1685.

In addition, the Shore Project Group requests that users grant the Computer Sciences Department rights to redistribute these changes.

 

What is New in Release 1.1.1

The Linux port has been brought up-to-date. Changes were small but rather pervasive. Recent releases have become more and more Solaris-specific. These updates should make Shore easier to port to other platforms.

Numerous other small bugs, mostly concerned with compiling and installing the software, have been repaired.

A rather obscure bug concerning SIX mode locking has been fixed in the lock manager (src/sm/lock.cc).

The binary release was built with the latest GNU compilers: gcc (and g++) version 2.7.2.3 and libg++ 2.7.2.

 

What was New in Release 1.1

 

Maintenance Fixes

Numerous bug fixes and performance improvements have been made throughout Shore, especially in the storage manager. In particular,

• Locks are automatically escalated from record to page to file level.
• Logging has been substantially improved to generate fewer log records.
• Bulk loading facilities have been improved.
• Allocation of space in files has been substantially improved.
• Numerous storage leaks and uninitialized variables discovered by Purify^TM have been fixed.
• More kinds of temporary files are available. The semantics have been cleaned up and bugs in logging have been fixed.
• The concurrency control and recovery code for btrees has been extensively revised more closely to follow Mohan's ARIES algorithms.
• Keys of all primitive types are now correctly handled (in a machine-independent manner) for btrees and for sorting.
• The #include files in the storage manager have been reorganized to reflect a cleaner layering and to improve compilation speed.
• Error messages from the diskrw process have been made more intelligible.
• A great deal of obsolete code has been removed.
• Most of the documents have been substantially rewritten.

 

Configuration

The system for configuring and building Shore from source has been substantially streamlined. We no longer use imake (from the X Window System) to generate Makefiles. Instead, Makefiles are generated from Imakefiles and configuration files in the config directory by a Perl script. In addition to eliminating the need to install imake, this change makes configuration considerably easier to understand and modify. Imakefiles and substantially shorter, and the generated Makefiles are both much shorter (up to 5 times shorter) and formatted more readably. All local configuration settings related to compiling are confined to a single file: config/config.h, and mostly to the first 160 or so lines. The only runtime configuration necessary is a one-line change to a single file (lib/options, which is normally installed as ~/.shoreconfig).

 

What was New in Release 1.0

 

General

• Ported to Solaris 2.5.
• The software builds under gcc 2.7.2 with the libg++ 2.7.1 library and include files.
• Numerous bugs were fixed.

 

Applications

• Applications linked with the no-debug libraries must be recompiled with the no-debug include files, because the two implementations of the return-code class are incompatible.
• Ref<Pool> can be used in applications.
• Enumerations can be used for index key types.
• Reference-counting on modules is enabled and under the control of the option oc_refcount. If you use this, use it with caution.

 

Object cache

No changes.

 

Shore server

See Storage manager.

 

Storage manager

• The format of logs and volumes has changed; data created under previous releases cannot be read by a Version 1.0 server.
• The SM programming interface is slightly changed to support multi-threaded transactions (in limited circumstances: see smthread_t(SSM) ),

  external two-phase commit (see transaction(SSM)), and faster record creation in a file in append-only context (see scan_file_i(SSM) ).

• The log can be a raw device. The SM can be configured to use a separate (diskrw) process for the log or to perform the log I/O locally (and synchronously). The SM layer has a new option sm_reformat_log to control reformatting the log, in the event that a raw device is used. The meaning of the option sm_logsize has changed. (See options(SVAS) for details.)
• More statistics are reported.
• Miscellaneous performance improvements were made (logging uninitialized data, append-only file population, store-descriptor cache expands as needed, lock manager simplified).
• The programming interface for threads has changed. The changes are incorporated into the example value-added server, and described in the companion document.