Unfortunately, the *roff `ad` request doesn't work like everyone seems to think it does. When invoked without an argument, it does not mean "restore whatever the previous adjustment mode was before I invoked `na` to turn off adjustment". Instead, it means "make sure lines are adjusted regardless of the previous adjustment mode, unless that mode was 'r' or 'c', in which case restore it." (These aren't, strictly speaking, "adjustment" modes since modes 'c' and 'r' don't adjust the widths of spaces.) No, this doesn't make any sense but it is nevertheless how AT&T *roff behaved starting somewhere between Sixth and Seventh Edition Unix, and GNU troff replicates the behavior.
Some devils at Murray Hill whispered into Joe Ossanna's ear and persuaded him that the alignment of text (to the left, right, or center of the output line) and the adjustment of spaces within that output line were not separate concerns. Is this an instance of brain damage that Ken Thompson brought back from his Berkeley sabbatical? Solution: Since AT&T troff also lacks a feature for hyphenless discretionary breaks (groff's `\%`), recast a sentence, set the example `PATH` on its own line, indent it as a sort of inline display, use the Ninth Edition Unix/groff extension `EX/EE` to set it in a constant-width typeface if possible, and break the line in a fortuitous place if it seems like it would overrun. For groff 1.25 (not the in-preparation 1.24), I plan to revise the behavior of the `ad` request to do what most of its users seem (reasonably) to expect, at the risk of angering expert users who knew about the `.j` register[1] but refused to use it. Here is a homework assignment for anyone inclined to argue with me. $ nroff <<EOF .ll 24n \" Short line length demonstrates adjustment. This text is aligned and adjusted per the formatter's defaults. .br .na This text is not adjusted. .br .ad This is left-aligned, adjusted text. .br .ad c This is centered text. .br .ad Predict the alignment and adjustment of this text. .br .ad r Now let's try right-alignment. .br .ad Predict the alignment and adjustment of this text. .pl \n(nlu EOF References: 1. https://lists.gnu.org/archive/html/groff/2024-06/msg00053.html 2. https://www.tuhs.org/pipermail/tuhs/2024-October/031012.html 3. https://savannah.gnu.org/bugs/?65954 [1] `.j` was a late addition to pre-Kernighan troff, appearing in Seventh Edition Unix (1979) but documented not CSTR #54 (1976), the Troff/Nroff User's Manual. Possibly it was contributed by hands other than Ossanna's after his untimely death. I wonder if this was done in realization of what a poor decision had been taken, a sort of partial repentance for an auto-craniotomy. --- doc/bash.1 | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/doc/bash.1 b/doc/bash.1 index 43c12b8d..2f0edf7d 100644 --- a/doc/bash.1 +++ b/doc/bash.1 @@ -2820,10 +2820,16 @@ .SS "Shell Variables" The default path is system-dependent, and is set by the administrator who installs .BR bash . -A common value is -.na -.Q /usr/local/bin:/usr/local/sbin:/usr/bin:/usr/sbin:/bin:/sbin . -.ad +A common value follows. +.RS +.RS 5n +.EX +/usr/local/bin:/usr/local/sbin:\c +.if \n(LL<78n .br +/usr/bin:/usr/sbin:/bin:/sbin +.EE +.RE +.RE .TP .B POSIXLY_CORRECT If this variable is in the environment when \fBbash\fP starts, the shell -- 2.30.2
signature.asc
Description: PGP signature
