Ingo Schwarze [Sat, 26 Aug 2017 15:55:46 +0000 (15:55 +0000)]
Do not fork and exec cmp(1); instead, simply fstat(2), mmap(2), and
compare the files directly, allowing a much stricter pledge(2), at
very little cost: merely 15 additional lines of very simple code.
Suggested by George Brown <321 dot george at gmail dot com> on misc@.
Ingo Schwarze [Sat, 26 Aug 2017 12:59:17 +0000 (12:59 +0000)]
No need to fork and exec rm(1) -rf, we know that we have exactly
one file and exactly one directory to remove. While here, increase
the size of the buffer such that the file name actually fits.
Minus 17 lines of code, no functional change.
Opportunity for simplification reported by George Brown <321 dot
george at gmail dot com> on misc@.
Ingo Schwarze [Wed, 23 Aug 2017 20:03:17 +0000 (20:03 +0000)]
Suppress spacing before certain kinds of lists.
One benefit is a reduced probablity that a blank appears between
a function name and the opening parenthesis introducing the arguments.
The heuristics isn't perfect and may occasionally suppress a blank
that wouldn't do harm.
Ingo Schwarze [Wed, 23 Aug 2017 10:50:15 +0000 (10:50 +0000)]
Switch ASCII rendering of the same mathematical symbols and greek
letters as in groff commit babca15f from trying to imitate the
characters' graphical shapes, which resulted in unintelligible
renderings in many cases, to transliterations conveying the characters'
meanings. One benefit is making these characters usable for portable
manual pages.
Ingo Schwarze [Mon, 21 Aug 2017 15:42:58 +0000 (15:42 +0000)]
When the stdout stream gets broken, there is no point in reading
any more input files, and it would be misleading to start a parser,
because that would show randomly truncated text.
Instead, print an error message and exit the program.
Issue found by Leah Neukirchen <leah at vuxu dot org>, who was
surprised to see half a manpage when her /tmp/ overflew.
Ingo Schwarze [Fri, 11 Aug 2017 16:56:21 +0000 (16:56 +0000)]
Make the "new sentence, new line" check stricter, allowing digits
in the last two letters of the last word of the sentence.
No false positives in base or Xenocara.
Suggested by and OK jmc@.
Ingo Schwarze [Wed, 9 Aug 2017 12:49:04 +0000 (12:49 +0000)]
Explain how to transform markup for the terminal when not using a
pager, and how to remove markup. Add related cross references.
While here, as suggested by jmc@, replace the excessive cross
references to the intro pages by a more relevant one to mandoc(1).
Triggered by a question from, using feedback from, and OK jmc@.
Ingo Schwarze [Wed, 2 Aug 2017 13:29:04 +0000 (13:29 +0000)]
No longer use names that only occur in the SYNOPSIS section as names
for man(1) lookup. For OpenBSD base and Xenocara, that functionality
was never intended to be required, and i just fixed the last handful
of offenders using it - not counting the horribly ill-designed
interfaces engine(3) and lh_new(3) which are impossible to properly
document in the first place.
Of course, apropos(1) and whatis(1) continue to use SYNOPSIS .Nm,
.Fn, and .Fo macros, so "man -k ENGINE_get_load_privkey_function"
still works.
This change also gets rid of a few bogus warnings "cross reference
to self" which actually are *not* to self, like in yp(8).
This former functionality was intended to help third-party software
in the ports tree and on non-OpenBSD systems containing manual pages
with incomplete or corrupt NAME sections. But it turned out it did
more harm than good, and caused more confusion than relief,
specifically for third party manuals and for maintainers of
mandoc-portable on other operating systems. So kill it.
Problems reported, among others, by Yuri Pankov (illumos).
Ignore explicitly specified negative column widths rather than
wrapping around to huge numbers and risking memory exhaustion;
fixes Debian ps(1). Bug reported by Dr. Markus Waldeck.
Fix an out of bounds read access to a constant array that caused
segfaults on certain hardened versions of glibc. Triggered by .sp
or blank lines right before .SS or .SH, or before the first .Sh.
Found the hard way by Dr. Markus Waldner on Debian
and by Leah Neukirchen on Void Linux.
When the mparse_alloc() prototype changed in read.c 1.179, updating
the enum constant was forgotten in this call. No functional change
since the mmsg argument is NULL anyway.
Found by florian@ with clang.
Cast the return value of chdir(2) to void. We already have a comment
above explaining to human auditors why no error checking is needed
here, so it's only fair to tell the compiler, too.
Worried compiler reported by Michael <Stapelberg at debian>.
Avoid cross reference to mandoc(1) because it may cause a message
on other operating systems where mandoc is not installed,
causing the test to fail in the portable test suite.
For -Tlint, put parser messages on stdout instead of stderr.
Originally, naddy@ requested this in 2011 (or maybe even earlier).
It was discussed with joerg@, kristaps@, naddy@, and espie@ in 2011,
and everybody agreed in principle, but it was postponed because
kristaps@ wanted to do some cleanup of the message system first.
Meanwhile, message infrastructure was improved about a dozen times...
This makes long, tedious commands like "mandoc -Tlint *.1 2>&1 | less"
unnecessary and allows simple ones like "man -l -Tlint *.1".
* clean catman.o which was missing
* delete WWW_OBJS which i don't use
* do not complain about *.out_markdown
* bump VERSION to 1.14.2 (not released yet)
Prefer arch-dependent over arch-independent pages if the name priority,
the section number, and all names match. Changes little on installed
systems except the ordering of apropos(1) results, because we install
base and Xenocara manuals in different trees, but fixes lookup of pages
like apm(4) vs. apm(4/amd64) in man.cgi(8).
Issue discovered by martian67 on freenode and reported via tj@.
Run the regression tests iteratively rather than recursively.
Simpler user interface, better summary output, simpler code,
and slightly faster.
Inspired by espie@'s request to avoid needless recursive make(1)
in the OpenBSD regression testing tree.
The only elements using floats are dt.It-tag,
so the only elements allowed to the right of them are dd.It-tag.
Fixes overlapping rendering after long tag lists without bodies,
for example in the Xenocara apm(4) manual. Issue reported by tj@.
In .Bl -tag and -hang, do not print a margin-left style attribute
for each individual item if the -width argument matches the default
of 6n. Suggested by Steffen Nurpmeso <steffen at sdaoden dot eu>
on <groff at GNU dot org> in April 2017.
If -column, -diag, -inset, -item, or -ohang lists have a -width,
don't just talk about ignoring it, actually do ignore it.
No change for terminal output, improves HTML output.
Improve rev. 1.250 further:
If a database is empty, we already don't write a database file and
delete an existing one. If none exists, that only means the database
was empty and remains so, which is not an error. So don't nag about
it, in particular because that would be a weekly(8) annoyance.
Issue reported by jmc@.
Explicitly initialize a variable where the compiler is (understandably)
unable to figure out that it is never used uninitialized.
While here, tweak the content of the variable to make its usage
easier to understand. No functional change.
Fix an assertion failure triggered by print_otag("sw+-l", NULL).
Even though we skip the style when the argument is NULL,
we must still consume the options.
Not found with afl(1), but during manual testing of the previous patch...
Handle .Bl -compact via CSS rather than writing individual style
attributes into .It blocks; suggested by Steffen Nurpmeso <steffen
at sdaoden dot eu> on <groff at GNU dot org> in April 2017.
Delete margin-bottom and margin-top style names and the 'v' argument
letter from print_otag() because they are no longer used.
Simplify by creating struct roff_node syntax tree nodes for tbl(7)
right from roff_parseln() rather than delegating to read.c,
similar to what i just did for eqn(7).
The interface function roff_span() becomes obsolete and is deleted,
the former interface function roff_addtbl() becomes static,
the interface functions tbl_read() and tbl_cdata() become void,
and minus twelve linus of code.
1. Eliminate struct eqn, instead use the existing members
of struct roff_node which is allocated for each equation anyway.
2. Do not keep a list of equation parsers, one parser is enough.
Minus fifty lines of code, no functional change.
Correctly handle horizontal spans at the beginning of rows,
fixing an assertion failure found by jsg@ with afl(1).
While here, also drop printing of whitespace in tbl_data()
which makes no difference because column positioning code
in term_tbl() already takes care of that.
Now that we have the -Wstyle message level, downgrade six warnings
that are not syntax mistakes and that do not cause wrong formatting
or content to style suggestions.
Also upgrade two warnings that may cause information loss to errors.
Fix operator precedence according to Brian W. Kernighan and Lorinda
L. Cherry, "Typesetting Mathematics - User's Guide (Second Edition)",
August 15, 1978, paragraph 23; swarm of bugs pointed out by bentley@.
The EQN_LISTONE box type is pointless.
Simplify by just using EQN_LIST with expectargs = 1.
Noticed while investigating a bug report from bentley@.
No functional change.
Messages of the -Wbase level now print STYLE:. Since this
causes horrible churn anyway, profit of the opportunity to stop
excessive testing, such that this is hopefully the last instance
of such churn. Consistently use OpenBSD RCS tags, blank .Os,
blank fourth .TH argument, and Mdocdate like everywhere else.
Use -Ios=OpenBSD for platform-independent predictable output.
Printing "BASE:" in messages about violations of base system conventions
is confusing, simply print "STYLE:", which is intuitive and does not
sound excessively alarming; suggested by jmc@, OK tedu@ jmc@.
If a single page references the same non-existent manual more than
once, print "(N times)" after the message "referenced manual not
found", to lessen the risk that people fix the first instance and
miss the others; jmc@ confirmed that this is useful.
When checking cross references with -Tlint, ultimately fall back to
looking in the current working directory. Not a security issue
because the files are never open(2)ed, only access(2)ed.
Requested by jmc@ and inspired by mdoclint(1).
This cannot be perfect because it only works for files having the
exact filename ./pagename.sec - mandoc has no way to figure out
which files might contain a manual for multiple names, or that files
in autohell might be called ./pagename.man.in instead, or which
subdirectories might contain additional source files. Also, it may
hide messages if you have bogus stuff lying around in the directory
where you run mandoc -Tlint. But jmc@ considers it important, and
good enough for everyday use.
Also avoid leaking the memory for the file name while here.
When checking the validity of cross references with -Tlint,
fall back from database search to file system search
just like man(1) does when looking up manuals.
This is not too expensive because on a system having up-to-date
mandoc.db(5) files, it only prolongs the time needed to check
*invalid* references - and you are not supposed to have many of
those, right? And on a system with missing or invalid mandoc.db(5)
files, spending a bit of time and warning loudly about the real
problem is also better than quickly issuing bogus warnings about
cross references that are actually valid.
Basic reporting of .Xrs to manual pages that don't exist
in the base system, inspired by mdoclint(1).
We are able to do this because (1) the -mdoc parser, the -Tlint validator,
and the man(1) manual page lookup code are all in the same program
and (2) the mandoc.db(5) database format allows fast lookup.
Feedback from, previous versions tested by, and OK jmc@.
A few features will be added to this in the tree, step by step.
Ingo Schwarze [Thu, 29 Jun 2017 16:31:15 +0000 (16:31 +0000)]
Skip whitespace at the beginning of eqn(7) nodes,
in particular ~ and ^ that misrendered;
found by bentley@ in glCopyTexSubImage1D(3); also affected
glAccum(3), glClipPlane(3), glDrawPixels(3), glEvalMesh(3), and others.
Ingo Schwarze [Wed, 28 Jun 2017 00:59:57 +0000 (00:59 +0000)]
Rewrite half of this, i was completely unaware how bad it was.
Remove several lies, lots of duplicate information,
and a lengthy discussion of features we don't support.
Clarify the wording in some places and make it more concise in others.
Delete examples from where they don't belong
and write a new EXAMPLES section from scratch.
Ingo Schwarze [Tue, 27 Jun 2017 18:25:02 +0000 (18:25 +0000)]
Implement spacing of columns as defined in the table layout;
this is for example used by lftp(1)
and, ironically, misused by our very own tbl(7) manual...