• SOC2010BenFiedler

Project Title

BSD-licensed Text-Processing Tools

Project Description

Merge the latest version of mdocml (ISC licensed) from the ports into the FreeBSD system base, and configure it as a replacement for the GNU roff (groff) tool set (this would aditionally involve a rewrite of man). There are FreeBSD replacements for sort and diff in ports (also from a GSoC project), but they are lacking features from their GNU counterparts. I will pick up and complete these, possibly borrowing/porting from one of the other BSD projects when possible. Benchmarking the new tools and ensuring they scale appropriately when compared to their GNU counterparts is highly desired, and I would like to try to confirm the mdocml author's assertion that it should be noticeably faster than groff.

Deliverables

A feature-complete (including wide character support) BSD-licensed replacement for:

• man, diff, sort, sdiff, diff3
• Integration of mdocml into FreeBSD base, completely replacing groff
• Integration of above tools into FreeBSD base
• Benchmark comparisons of above tools to previous implementations

General TODOs

• Identify use of man, diff, sort, sdiff, or diff3 in make scripts, build world, etc.
• Performance analysis toolset

Project Schedule / Milestones

May 24 - May 26: Licensing/Cost Analysis

• Identify existing tools with a BSD-compatible license (ISC, 2/3/4 Clause BSD, Public Domain?). Contact respective authors about possibility of changing existing code to 2 Clause BSD licensing.
• Brief analysis for above candidates of effort required to port existing code to FreeBSD versus scratch re-write.

June 1 - 12: diff

• (vacation ends)

• complete diff replacement/port

June 13 - June 15: diff Testing

• Verify feature compatibility
• Exhaustive debugging
• Performance-measuring, optimization if necessary

June 16 - June 21: sdiff

• Complete sdiff replacement/port

June 21-25: sdiff Testing

• Verify feature compatibility
• Exhaustive testing (mergemaster)
• Performance-measuring, optimization if necessary

June 26 - July 16: mdocml

• Merge mdocml (mandoc) into base.

• Work with Ulrich Spörlein to prevent duplicate work

• Identify and replace complement scripts with option of compiling man pages with mandoc tools

• Find and recompose any pages not compatible with mandoc

• GSoC Mid-term *

July 17 - July 19: man

• Complement http://people.freebsd.org/~gordon/man.sh (broken link) man replacement

• Ensure compatibility with mandoc

July 20 - July 26: sort

• Complete sort replacement/port

July 27 - July 30: '''sort''' Testing

• Verify feature compatibility
• Regression tests
• Performance-measuring, optimization if necessary

July 31 - August 7: diff3

• Complete diff3 replacement/port

August 7 - August 9: '''diff3''' Testing

• Verify feature compatibility
• Regression tests
• Performance-measuring, optimization if necessary

August 9 - August 16: Closing Time

• Documentation
• Compile performance presentation
• Test, test, test

==Status Report==

diff

sdiff

Base version ported from OpenBSD

diff3

Base version ported from OpenBSD

mandoc

Complemented Ulrich Spoerlein's progress: system makefiles have been modified to build mandoc instead of groff and use Gordon Tetlow's modified man.sh when WITH_MANDOC specified in /etc/make.conf .

Remaining items TODO:

• Complete patch system for changing mandoc-incompatible man pages.
• Set up a ports package for groff (textproc/groff ?)

• Identify tools with remaining groff dependencies (catman, whatis, ...) (looks like Gordon has since done work on this)

  • Finish gordon's man.sh by removing dependencies on groff tools: (eqn, pic, refer)
• Identify remaining mandoc-incompatible man pages. Patch.