releasenotes last touched in Par 1.53.0 last meaningful change in Par 1.53.0 Copyright 1993, 1996, 2000, 2001, 2020 Adam M. Costello Each entry below describes changes since the previous version. Par 1.53.0 released 2020-Mar-14 Fixed the following bugs: An unintended bad interaction between and . The specification was inconsistent. Although it said that the lines inserted by the feature were vacant, the feature could interpret the quote character of inserted lines as a repeat character, in which case the lines were not vacant (according to the definition), and more quote/repeat characters would be inserted to extend the line to the full width, which no one would ever want. The definition of "bodiless line" is revised so that lines inserted by the feature are excluded from consideration by the feature. A printf format string mismatch (ptrdiff_t vs. %d), reported by Quentin Barnes (qbarnes at gmail.com). protoMakefile's clean target removed par rather than par$E, reported by George V. Reilly (george at reilly.org). Added the following features: Locale support for single-byte character sets. The previous verson, 1.52, attempted to do this in one line of code (setlocale()), but it's not that simple. Versions before 1.52 gave no thought to character sets other than US-ASCII and did not handle them entirely correctly. Calling setlocale() in version 1.52 corrected some flaws but not all, and created one new flaw. This version and the previous two all have the same character handling in the "C" locale and when processing US-ASCII text; differences arise only when processing non-US-ASCII text in another locale. In versions before 1.52 the implementation assumed that "uppercase letter" means only the 26 characters A-Z and that "lowercase letter" means only the 26 characters a-z, and the specification assumed that every letter is either upper case or lower case. These assumptions hold for US-ASCII, but not for most other character sets; therefore versions before 1.52 did not really support non-US-ASCII text. Version 1.52, by calling setlocale(), relaxed the three assumptions in the implementation but not the assumption in the specification, and inadvertantly caused the implementation to deviate from the specification by converting all space characters to spaces rather than converting only the specified white characters, which is not necessarily what you want (for example, you probably don't want no-break space converted to space, but some locales on some platforms have been known to classify no-break space as white-space, like FreeBSD did in 2004). This version fixes the specification by introducing the concept of "neither-case letter" and redefining "alphanumeric character" to include it, and fixes the implementation to convert only the white characters to spaces. It also makes the white characters configurable so that the version 1.52 space-conversion behavior can be restored if desired (W=_S). Note that Par still assumes constant display width per byte, which is not true for UTF-8 nor ISO-2022-*, but is true for most single-byte charsets and for EUC-* text without the single-shift-in/out codes. The W option, for configuring the set of white characters. The Z option, for configuring the set of terminal characters. The _@ escape sequence (neither-case letter) in charset syntax. The _S escape sequence (any space) in charset syntax. Added _@ and apostrophe to the set of body characters in PARINIT. Add #includes for whatever interfaces are used rather than depending on included files to #include other files. In protoMakefile introduced CPPFLAGS and CFLAGS (default empty) into the default definition of CC. Added a test-par script and a test target in protoMakefile. Fixed a misspelling (preceed => precede) in par.doc and par.1. Changed the top-of-file comment convention so that every release won't need to touch every file (this is the last time). Updated the author contact information. Clarified the license and added the MIT License as an alternative. Adopted a modern version number format, with minor and patch numbers as dot-separated integers rather than adjacent single digits. Par 1.52 released 2001-Apr-29 Fixed a portability problem regarding unsigned char versus char. (This was a potential problem only on platforms where char is a non-two's-complement signed type, which is very few, if any.) Added a call to setlocale(), for 8-bit character support, requested by Alex Postnikov (apost at math.berkeley.edu) and Oliver Fromme (olli at secnetix.de). Added some unnecessary parentheses, braces, and initialization, to quiet the more aggressive compiler warnings. (This was suggested back in 1993 by Thomas E. Dickey (dickey at software.org) but I refused. I guess my attitude has changed.) Changed the EXAMPLES section of par.1 to use the constant-width font rather than constant-spaced Roman, suggested by Carl Lowenstein (cdl at mpl.ucsd.edu). Reorganized releasenotes to put the most recent version first, and rearranged the email addresses to confuse address harvesters. Par 1.51 released 2000-Feb-24 Updated the author contact information. Relaxed the copyright restrictions. Added comma to the recommended set of body characters in PARINIT. Par 1.50 released 1996-Feb-11 Added the following features: The T option, suggested by Jules Junker (jjunker at eos.hitc.com). The b option. Loosened the terms of the Copyright. Changed protoMakefile: Added the E and O macros, suggested by Peter A. Getterman (pgutterman at worldbank.org). Added sensible default values for all macros. Added question-mark to the recommended set of body characters. Restored documentation of the number-without-a-letter shortcut command-line option, which got lost between versions 1.00 and 1.10. Its absence was noted by David W. Sanderson (dws at ssec.wisc.edu). Added .exrc examples, suggested by Alex Matulich (matuli_a at marlin.navsea.navy.mil), though I used my own .exrc. Added the Apologies section to par.doc and par.1. Made many other tweeks to the documentation. Removed the useless synopsis from the runtime help message. Par 1.41 released 1993-Oct-31 Moved the former Release Notes section of par.doc into its own file. Fixed the following bugs: A couple of inconsistencies of style in the code, pointed out by Thomas E. Dickey (dickey at software.org). If and were 1 and a one-line paragraph beginning with quote characters was reformatted into a multi-line paragraph, the new lines began with spaces instead of quote characters. This was because the policy for copying prefixes used the fallback prelen rather than the augmented version of it used for computing the default value of . Now both use the same formula. Thanks go to Steven King (king at rtsg.mot.com) for reporting the problem. If the t option were given without a number, it was unset instead of being set to 1. Added the following feature: The E option, suggested by Alex Matulich (alex at bilver.oau.org). Altered the terminology: Added the term "augmented fallback prelen". Added to the Rights and Responsibilities section of par.doc a guideline for handling version numbers in patched versions of Par. Added a useful suggestion to protoMakefile. Par 1.40 released 1993-Oct-10 Fixed the following bugs: The phrase "containing at least two lines" was left out of the definition of "vacant line" in par.doc and par.1, although the code implemented the correct definition. (The phrase now appears in the definition of "order bodiless line".) There was still a flaw in the design of the feature. If two adjacent lines had quoteprefixes neither of which was a prefix of the other, no line would be inserted between them, possibly causing other inserted lines not to be vacant. Now it should be true that every line inserted or truncated by the feature will turn out to be vacant. When began affecting the default value of (as of Par 1.20), it should also have begun affecting the choice of source of characters for the suffixes of lines in OPs. Added the following features: The i option, suggested by Alan Barrett (barrett at ee.und.ac.za). The e option, inspired by a suggestion from Tim Pointing (tim at ben.dciem.dnd.ca). The r option. The p, s, and w options are now accepted without numbers. par no longer gives up so easily when choosing default values for and for an IP with less than +2 lines, nor when choosing the source of characters for the prefix and suffix of the th output line when > and <= . These are incompatible changes, but I cannot imagine anyone preferring the old behavior. Altered the terminology: Added the terms "bodiless line" (a generalization of "vacant line"), "fallback prelen", and "fallback suflen". Made miscellaneous changes to the documentation, including the addition of the Quick Start section. Par 1.32 released 1993-Sep-13 Fixed the following bugs: par could crash when the + or - operator was used with the B, P, and Q options. Thanks go to Andrew Cashin (splat at deakin.oz.au) for reporting this. If were 1 and two adjacent lines had different quoteprefixes, one of which was a prefix of the other, and only the line with the shorter quoteprefix contained a non-quote character, then nothing would be altered. According to par.doc 1.31, this was correct, but since neither line is vacant, I consider this a bug in the design of the feature. Now the longer quoteprefix will be truncated to match the shorter one, and will therefore be vacant. Thanks go to Andrew Cashin (splat at deakin.oz.au) for asking about this. Made slight changes to the documentation. Par 1.31 released 1993-Sep-07 The version number is 1.31 rather than 1.40 because all added features are really just enhancements of existing features. Fixed the following bug: In par.doc, in the example of a paragraph produced by a greedy algorithm, the word "establish" appeared twice in a row. Thanks go to Daniel Kussendrager (daniel at astro.rug.nl) for first pointing this out. (The example is now even better because the paragraph looks even worse than before.) Added the following features: A usage message to accompany command line or environment variable syntax errors, first suggested by Karl Stiefvater (qarl at ecl.wustl.edu). The help and c options. The B, P, and Q options, which render PARBODY, PARPROTECT, and PARQUOTE no longer necessary. They are retained, though, for compatibility and convenience. The _b, _q, and _Q escape sequences for charset syntax. Added the term "charset syntax". Isolated the character set code in charset.c and charset.h. Par 1.30 released 1993-Aug-18 Since Par 1.20 was posted to comp.sources.misc, I have made only backward-compatible changes in Par 1.30. Fixed the following bugs: One wrong word in par.c sometimes caused par to crash. Thanks go to Contr Karl Vogel (vogelke at c-17igp.wpafb.af.mil) for sending me an input file that caused a crash. Too-long words were chopped up before the first word in a paragraph was expanded to include initial spaces, allowing impossibility #1 to occur. The order of the two operations has been reversed. Thanks go to Andrew Cashin (splat at deakin.oz.au) for reporting the error message. Added the following features: The g option, motivated by suggestions from several people. The q option, inspired by a suggestion from Andrew Cashin (splat at deakin.oz.au). The R option (my attempt to squash a bad idea from Par 1.00). The PARQUOTE environment variable (comes with the q option). The PARPROTECT environment variable, inspired by a suggestion from Dennis Flaherty (dennisf at se01.elk.miles.com). Altered the terminology: Several terms have been added, and the meaning of some terms has been slightly modified. This is a change in the language used to describe par's behavior, not a change in par's actual behavior. Added a clean target to protoMakefile, suggested by Hal Jespersen (hlj at posix.com). Par 1.20 released 1993-Aug-10 Since Par 1.10 was distributed to no one, I've made some more incompatible changes in Par 1.20. Added the following features: The d option. Paragraphs are now separated by vacant lines, not just blank lines. now affects not only but also . Par 1.10 released 1993-Aug-02 Fixed the following bugs: In reformat.c I used sprintf() but forgot to #include . I forgot to verify that > + . The first word of a paragraph was expanded to include initial white characters, not just spaces, contrary to par.doc. Some invalid options were not complained about. NUL characters in the input were not handled. A pointer foul-up in freelines() in par.c could cause a crash. Added the following features: The f, j, and t options. The PARBODY environment variable. Multiple options may be concatenated into a single argument. Removed the m option: Its function is better performed by the f and t options. Normally I would avoid making incompatible changes, unless I were doing a complete overhaul of the whole program, in which case I'd make the version number 2.00 to alert users to possible incompatibilities. However, in this particular instance I allowed an incompatibility in a minor upgrade because version 1.00 was distributed to only four people. Changed the handling of white characters: par now changes all of them (except newlines) to spaces as they are read. This is another incompatible change, excused for the same reason. Made all error messages begin with "par error:". Par 1.00 released 1993-Jul-25 The first release.