Completely removed the html5lib sanitizer.
Fixed the string-handling to work in both
Ruby 1.8.x and 1.9.2. There are still,
inexplicably, two functional tests that
fail. But the rest seems to work quite well.
Touch the tmp/restart.txt file, when
rotatingthe log files. Otherwise, multiple
workers may try to rotate the log files
at the same time, with sub-optimal results.
Also, an aesthetic tweak to the url_generator.
% rake upgrade_instiki
fixes some potential problems in the database column types.
Revision content can now be up to 16MB.
Under MySQL, the previous limit was 64KB.
Page names can now be up to 255 bytes.
Under MySQL, the previous limit was 60 bytes.
Additional CSS styles can now be up to 64KB.
Under MySQL, the previous limit was 255 bytes.
Thanks to Andrew Stacey for reporting these.
Apply the same methodology, as in Revision 432,
to the category chunk-handler. This completes
the replacement of all the code that looks like
if string.is_utf8?
do something
else
complain
end
with code that looks like
string.purify
do something
Previously, if the user tried to submit content which was
malformed utf-8, Instiki would complain loudly to him.
A slightly more user-friendly approach was suggested by
the latest Rails 2.3.4, and a conversation with Sam Ruby
(who suggested some improvements).
Now, instead of complaining, we remove the offending bytes,
leaving a well-formed utf-8 string, which we pretend is what
the user meant to submit.
Should be to the published action. This
didn't work right for inter-web links.
(Reported by Mike Shulman)
Also, change some .length's to .size's
(for Andrew Stacey)
Refactor the upgrade_instiki rake task.
Based on the (very nice) JHerdman's
64d305f2a8
but defaults to 'production' environment, instead.
Instiki users don't know about production/development/test.
Instiki defaults to 'production'. So should its associated rake tasks.
Added the ability to rename existing pages.
[[!redirects Some Page Name]] redirects Wikilinks [[Some Page Name]] to
the current page (assuming "Some Page Name" does not exist).
Real pages trump redirects (though this may change, depending on
user feedback).
From Jason Blevins:
Create a "History" page for each wiki page.
Link to it, and to the "Diff" page from "Recently Revised".
Also, correct a bug in listing/deleting links to uploaded
video and audio files.
Using <object> and <embed> were forbidden for obvious
security reasons. Instiki now permits embedding video
via the HTML5 <video> element (Ogg/Theora encoded videos
only, with .ogg or .ogv extensions). You can even upload
videos with
[[foo.ogg:video]]
Instiki now support x-sendfile. See the Proxying page for
configuring Apache (with the x-sendfile module). Lighttpd
should work similarly.
Update Rails to latest Edge (hopefully converging on RC2!).
Instiki now runs on the Rails 2.3.0 Candidate Release.
Among other improvements, this means that it now
automagically selects between WEBrick and Mongrel.
Just run
./instiki --daemon
On Webs with file uploads enabled, uploaded files were stored
(in version 0.16.1 and earlier) in the public/ directory.
This was a security threat. A miscreant could upload a .html file.
When a user clicked on the link to the file, it was opened (unsanitized)
in the browser.
As of version 0.16.2, uploaded files are stored in the webs/
directory. Now, when the user clicks on the link, the file is sent
with the
Content-Disposition: attachment
header set, which causes the file to be downloaded, rather than opened
in the browser. As always, files downloaded from the internets should be
treated with caution. At least, this way, they are not aoutomatically
opened in the browser.
To move your existing uploaded files to the new location, do a
rake upgrade_instiki
When a Web uses one of the Markdown Text Filters, and you export
all the pages as a zip file, you'd like the MathML and SVG to
render when the pages are viewed locally. This means saving them
with a .xhtml extension. Users of non-XHTML-capable browsers or
Textile users should still get .html files.
Ruby's String.sub!(pattern, replacement) routine is fundamentally
broken. But the block version works fine.
Using the broken routine in the Chunk handler was a subtle mistake.