Michael Smith [Sun, 23 Mar 2008 04:37:37 +0000 (04:37 +0000)]
use XSLT method to determine version number from VERSION file
(instead of grep hack); also, use w3m as the default browser for
generating plain-text output from HTML (with GC_NPROCS=1 to
prevent it from hanging under OSX/Darwin)
Michael Smith [Sun, 23 Mar 2008 04:32:34 +0000 (04:32 +0000)]
added "exslt" namespace binding to html/docbook.xsl file (in
addition to existing "exsl" binding. reason is because lack of it
seems to cause processing problems when using the profiled
version of the stylsheet
Michael Smith [Fri, 21 Mar 2008 05:12:40 +0000 (05:12 +0000)]
in install.sh, replaced use of non-portable "readlink -f" with
combination of cd/dirname/pwd that has the same effect. This is a
partial fix for #1735710. Thanks to Ivan Vilata i Balaguer.
Michael Smith [Mon, 10 Mar 2008 03:53:20 +0000 (03:53 +0000)]
set output of the literal element in monospace. this causes all
inline monospace instances in the git man pages to be set in
monospace (since DocBook XML source for git docs is generated with
asciidoc and asciidoc consistently outputs only <literal> for
inline monospace (not <command> or <code> or anything else).
Of course this only affects non-TTY output...
Michael Smith [Mon, 10 Mar 2008 02:56:27 +0000 (02:56 +0000)]
don't bother using a custom register to store the previous
font-family value when setting blocks of text in code font; just
use \F[] .fam with no arg to switch back
Michael Smith [Sat, 8 Mar 2008 17:50:28 +0000 (17:50 +0000)]
Added new param man.base.url.for.relative.links .. specifies a
base URL for relative links (for ulink, @xlink:href, imagedata,
audiodata, videodata) shown in the generated NOTES section of
man-page output. The value of man.base.url.for.relative.links is
prepended to any relative URI that is a value of ulink url,
xlink:href, or fileref attribute.
If you use relative URIs in link sources in your DocBook refentry
source, and you leave man.base.url.for.relative.links unset, the
relative links will appear "as is" in the NOTES section of any
man-page output generated from your source. That's probably not
what you want, because such relative links are only usable in the
context of HTML output. So, to make the links meaningful and
usable in the context of man-page output, set a value for
man.base.url.for.relative.links that points
to the online version of HTML output generated from your DocBook
refentry source. For example:
Michael Smith [Sat, 8 Mar 2008 04:28:17 +0000 (04:28 +0000)]
If a source refentry contains a Documentation or DOCUMENTATION
section, don't report it as having missing AUTHOR information.
Also, if missing a contrib/personblurb for a person or org, report
pointers to http://docbook.sf.net/el/personblurb and to
http://docbook.sf.net/el/contrib
Michael Smith [Fri, 7 Mar 2008 20:42:35 +0000 (20:42 +0000)]
If we encounter an author|editor|othercredit instance that lacks a
personblurb or contrib, report it to the user (because that means
we have no information about that author|editor|othercredit to
display in the generated AUTHOR|AUTHORS section...)
Michael Smith [Fri, 7 Mar 2008 09:18:28 +0000 (09:18 +0000)]
refined handling of cases where refentry "source" or "manual"
metadata is missing or when we use fallback content instead. We
now report a Warning if we use fallback content.
Michael Smith [Wed, 5 Mar 2008 15:16:10 +0000 (15:16 +0000)]
Made reporting of missing refentry metadata more quiet:
- we no longer report anything if usable-but-not-preferred
metadata is found; we just quietly use whatever we manage to
find
- we now only report missing "source" metadata if the refentry
is missing BOTH "source name" and "version" metadata; if it
has one but not the other, we use whichever one it has and
don't report anything as missing
The above changes were made because testing with some "real world"
source reveals that some authors are intentionally choosing to use
"non preferred" markup for some metadata, and also choosing to
omit "source name" or "version" metadata in there DocBook XML. So
it does no good to give them pedantic reminders about what they
already know...
Also, changed code to cause "fixme" text to be inserted in output
in particular cases:
- if we can't manage to find any "source" metadata at all, we
now put fixme text into the output
- if we can't manage to find any "manual" metadata a all, we
now put fixme text into the output
The "source" and "manual" metadata is necessary information, so
buy putting the fixme stuff in the output, we alert users to the
need problem of it being missing.
Michael Smith [Wed, 5 Mar 2008 10:19:29 +0000 (10:19 +0000)]
for a para or simpara that is the first child of a callout,
suppress the .sp or .PP that would normally be output (because in
those cases, the output goes into a table cell, and the .sp or .PP
markup causes a spurious linebreak before it when displayed
Michael Smith [Wed, 5 Mar 2008 10:13:59 +0000 (10:13 +0000)]
Added support for rendering co callouts and calloutlist instances.
So you can now use simple callouts -- marking up programlisting
and such with co instances -- and have the callouts displayed in
man-page output. ("simple callouts" means using co@id and
callout@arearefs pointing to co@id instances; in man/roff output,
we can't/don't support markup that uses areaset and area)
Michael Smith [Tue, 4 Mar 2008 10:40:21 +0000 (10:40 +0000)]
When generating manpages output, we no longer report anything if
the refentry source is missing date or pubdate content. In
practice, many users intentionally omit the date from the source
because they explicitly want it to be generated.
Michael Smith [Tue, 4 Mar 2008 07:07:43 +0000 (07:07 +0000)]
Made refinements/fixes to output of orderedlist and itemizedlist
-- in part, to get mysql man pages to display correctly. This
change causes a "\c" continuation marker to be added between
listitem markers and contents (to ensure that the content remains
on the same line as the marker when displayed)
Michael Smith [Tue, 4 Mar 2008 04:20:37 +0000 (04:20 +0000)]
Deal with case of verbatim environments that have a linebreak
after the opening tag. Assumption is that users generally don't
want that linebreak to appear in output, so we do some groff
hackery to mess with vertical spacing and close the space.
Michael Smith [Tue, 4 Mar 2008 03:27:47 +0000 (03:27 +0000)]
convert no-break space character to groff "\ \&" (instead of just
"\ "). the reason is that if a space occurs at the end of a line,
our processing causes it to be eaten. a real-world case of this is
the mysql(1) man page. appending the "\&" prevents that
Mauritz Jeanson [Mon, 3 Mar 2008 21:19:02 +0000 (21:19 +0000)]
Copy over the libxslt README too (currently there is no other documentation).
Do not delete the extensions directory when cleaning; only delete (non-dot) files in it.
Mauritz Jeanson [Mon, 3 Mar 2008 15:59:40 +0000 (15:59 +0000)]
Changed the name of the script file back to "xslt.py". Using "libxslt.py"
leads to a name clash since the script contains a "import libxslt" statement.
Michael Smith [Mon, 3 Mar 2008 05:07:39 +0000 (05:07 +0000)]
output "sp" before simpara output, not after it (outputting it
after results in undesirable whitespace in particular cases; for
example, in the hg/mercurial docs
Michael Smith [Sat, 1 Mar 2008 08:11:54 +0000 (08:11 +0000)]
Reworked output for admonitions (caution, important, note, tip,
warning). In TTY output, admonitions now get indented. In non-TTY
output, a colored marker (yellow) is displayed next to them.