Add mode for -Tlocale. This mode, with this commit, behaves exactly
like -Tascii. While adding this, inline term_alloc() (was a one-liner),
remove some switches around the terminal encoding for the symbol table
(unnecessary), and split out ascii_alloc() into ascii_init(), which is
also called from locale_init().
In tbl layouts, we puked if a space didn't followed a vertical bar
(found by Yuri Pankov). This was due to looking for modifiers for the
vertical bar. This has been fixed, along with other special-key layout
types.
Documentation: note COMPATIBILITY of -Tascii `?' printing in mandoc.1
and remove some long-fixed notes in sthe same section. Also, add an
`Lb' for the mandoc library to mandoc.3 (noted by Sascha Wildner).
Flip on printing `?' at Unicode codepoints in -Tascii, -Tpdf, and -Tps.
The reasoning behind printing SOMETHING at a Unicode codepoint is
because the input is not "wrong" (we suppress printing of "wrong"
things). It's just that ASCII can't handle it.
Cleanups in -T[x]html: make html_idcat() use the buffer and be called
bufcat_id(), then collapse it into a little function without so much
crap. Next, make bufinit() only be called when we really need to do so,
and not simply before pre/post calls.
Clean-ups in -T[x]html: inline print_num(), as it was just a single
conditional; same for print_xmltype() and print_doctype(), same reason;
make bufncat() be static, as it was only being called from html.c;
have bufcat() simply call through to strlcat(). Finally, assert()
whenever we truncate.
Also rename buffmt() -> bufcat_fmt() to differentiate from buffmt_man et
al., which do not concatenate.
Clean up -T[x]html by using a table instead of a switch statement for
the roff units. Also remove a comment about CSS and number types (they
all accept decimal numbers).
Remove function calls to res() and so forth in term_word(). These were
only used once and simply bloated the binary. Also fix mchars_num2char
to correctly render the character instead of using atoi(). This makes
the conversation more strict, but it's more correct.
Move struct termp_ps into term_ps.c; remove the engine union in struct termp,
which only held one entry; finally (as per the first), make "ps" member into a
pointer managed by term_ps.c. This frees up a nice chunk of memory during
run-time and in the binary.
Make character engine (-Tascii, -Tpdf, -Tps) ready for Unicode: make buffer
consist of type "int". This will take more work (especially in encode and
friends), but this is a strong start. This commit also consists of some
harmless lint fixes.
Give -Thtml and -Txhtml the gift of recognising escapes when calculating
widths (e.g., `Bl -tag -width "\s[blahblah]bar"). This has long since
been done for -Tascii but escaped noticed with -T[x]html.
Add configurations (`Cd') to mandoc-db mining.
Also put some notes into index.sgml to the effect that mandoc-db exists,
but is not linked to the build.
Back out stripping of non-predef and non-special escape sequences from
input (this is not yet possible with mandoc_escape(), which depends on
nil-terminated strings).
Move "chars" interface out of out.h and into mandoc.h. This doesn't
change any code but for renaming functions and types to be consistent
with other mandoc.h stuff. The reason for moving into libmandoc is that
the rendering of special characters is part of mandoc itself---not an
external part. From mandoc(1)'s perspective, this changes nothing, but
for other utilities, it's important to have these part of libmandoc.
Note this isn't documented [yet] in mandoc.3 because there are some
parts I'd like to change around beforehand.
Closing delimiters only suppress spacing when they follow something.
Fixing a regression introduced in rev. 1.105.
ok and prodding for comments kristaps@.
Make the `Nm' -Thtml attribute be min-width instead of width. This is a
quick fix for, say, rc.d(8) in OpenBSD, which has nested macros on the
`Nm' SYNOPSIS line that were skipped over by the length calculator. This
should [maybe?] be a recursive length check, but still it'd need to be
a min-width to accomodate for (say) `Qq' and the like printing excess
characters post-length-calculation.
Add \*(Ai (ANSI) and \*(Px (POSIX) predefined strings, which are part of
groff's tmac.doc package. Originally noted by Matthew Dempsky.
Feedback by Jason McIntyre, joerg@, and schwarze@. Also add some
documentation about predefined strings, tweaked by schwarze@.
Clean up parsing of delimiters in -mdoc. First, remove the "dowarn"
variable from mandoc_getarg() so that it prints the warning every time.
Then, remove the warning from args_checkpunct(). This way, warnings
are being posted at the correct time. This makes the flag argument to
mdoc_zargs() superfluous, so make it be zero when it's invoked. Finally,
move the args() flags into mdoc_argv.c and make them enums.
The semantics of .Bk was described incorrectly
for the case of multiple sibling macros on a single input line.
Issue found investigating a question from sobrado@.
"I like this diff" kristaps@
Check in fix to roff conditional if/else stack running out of space.
This transforms the stack pop to occur prior to body execution, instead
of afterward. Floated to tech@ without response, but it makes sense
that this is alright and doesn't cause problems during extensive
testing.
Remove the warning for empty bodies of `Sh', `Ss', `SH', and `SS'. This
prompted by a TODO by schwarze@, originally from Gleydson Soares, that
an empty `SS' was raising an error (it hasn't for some time). It makes
sense these shouldn't warn, as omitting their contents doesn't change
anything in the structure of the document (groff and mandoc specifically
account for the whitespace between empty sections).
This doesn't change any manuals, which only refer to the line arguments
(or possibly next-line, in the case of man(7) syntax).
Have mandoc-db accumulate manual page descriptions (`Nd' in -mdoc parlance)
in the index. This allows, with both the btree and index, full emulation
of apropos(1) and other goodies.
\# Everything up to and including the next newline is
ignored. This is interpreted in copy mode. This is like \"
except that the terminating newline is ignored as well.
Use dbt_xxxx functions to stash both filename and manual section in the
value part of the index. This is the actual manual section---before,
mandoc.cgi was relying on the file suffix, but this can be (e.g.) .man or
whatnot. This is The Correct Way (tm).
Remove a2roffdeco() and mandoc_special() functions and replace them with
a public (mandoc.h) function mandoc_escape(), which merges the
functionality of both prior functions.
Reason: code duplication. The a2roffdeco() and mandoc_special()
functions were pretty much the same thing and both quite complex. This
allows one function to receive improvements in (e.g.) subexpression
handling and performance, instead of having to replicate functionality.
As such, the mandoc_escape() function already handles a superset of the
escapes handled in previous versions and has improvements in performance
(using strcspn(), for example) and reliable handling of subexpressions.
This code Works For Me, but may need work to catch any regressions.
Since the benefits are great (leaner code, simpler API), I'd rather have
it in-tree than floating as a patch.
First, properly escape periods with \&. Then consistently refer to
non-macro lines as "text lines" (after defining them as such). Clean up
notion of font scope. Note comma usage in multiple `Nm's. Note
ordering of SYNOPSIS section stuff. Add some bits about DESCRIPTION
syntax (brief description, then args). Note that `Bl' can also contain
HEAD. Finally, fix some `Cm' entries to be properly `Ar'.
On .de macro lines, after the macro name, space and tab are equivalent.
Bug reported by Tristan dot LeGuern at gmail dot com in fvwm2(1).
tweaks and ok kristaps@; earlier version looked good to espie@ as well
Make `Rv' do the Right Thing regarding commas and "and" when listing its
output. This also makes it easier to read. While here, remove
superfluous assignment to local variables for this and `Mt'.
git.ckatri.com / mandoc.git / log