flagged string even if the locale causes it to generate utf8 output,
so make sure to let perl know it should be handled as utf8. Also,
the optimised version used for standard time formats won't work if the
user has changed locale, so drop it. Thanks, Faidon Liambotis.
* Fix re-encoding of the comments field to utf8 if a commit fails
due to a conflict. Thanks, Faidon Liambotis.
* Let svn know that commits have utf8 commit messages. Thanks, Faidon
Liambotis.
* Add insane double encode/decode to utf8 around call to markdown.
This works around a truely strange bug, which is apparently a bug in
perl, which I lack space to describe here (see t/crazy-badass-perl-bug.t)
- The serious problem is that writefile() saves the content in undecoded
format in CGI, hence all (non-ascii) chars are corrupted permanently.
- Comment field in edit page is not decoded and all the non-ascii chars in
this field are corrupted as the result. We should decode it as we do it
for the content.
layer, which led to lots of problems; make it force read files as utf-8.
Closes: #373203
* writefile() likewise needs to use the utf8 output layer.
* Remove the -CSD from ikiwiki's hashbang since it's useless to have it
there.
* Revert some of the decode_utf8 changes in CGI.pm that seem unnecessary
given the readfile fix.
* Add utf-8 testcases for readfile and htmlize.
* When inlining a page in another one, links from the inlined page are now
expanded the same as they are when rendering the inlined page as a
standalone page. So rather than being expanded from the POV of the
inlining page, they are expanded from the POV of the inlined page.
For example, a link from blog/foo to "bar" will now link to blog/bar
if it exists. Previously this needed to be a link explicitly to
"blog/bar"; such links will also continue to work.
(This was slightly complex to do as the link still has to be constructed
relative to the inlining page.)
* Add a html validity check to the test suite, using the wdg-html-validator,
if available.
* Make the html valid when there is nothing in the actions list by adding an
empty <li> to the end of it.
* Reordered some function call parameters for consistency.
for other types of search engine plugins if wanted, and also opening up a
lot of new possibilities for other kinds of plugins later
some notable changes along the way:
- lots of new hook types: cgi, render, delete
- wrapper files fixed to support config strings with newlines in them
- HEADERCONTENT in page template useful for plugins. Probably needs to be
expanded to more such for other places plugins might want to add content.
- remove unnecessary wrappers field from config info stored in wrappers
wikilinks since the urls should work now in more situations
- drop --limit from svn log run, since a) it needs a fairly new svn and
b) in some cases, it would limit it to too few entries to display the
requested number of changes
- Use driver:DB_File and not driver:db_file for better compatability with
old versions of CGI::Session.
- Note that HTML::Template 3.02.02 is needed.
so I don't need to maintain two copies anymore.
You might also want to remove the files provided in the basewiki underlay
from your wiki, if you have not created custom local versions of them, so
that these pages will be automatically updated in future ikiwiki upgrades.
- escape & in urls (also clean up cgi url generation)
- since markdown wraps inlined pages in <p></p>, close and re-open
the paragraph tags when generating the embedded html
- added XHTML 1.0 doctypes to templates
- fixed <hr /> and <br /> in templates
- add an alt attribute to inline images, based on the WikiLink to the
image. Allows things like [[my_image|img.png]] to customise alt text.
in filenames. Now converts to valid filenames automatically.
Note, need to --refresh your wiki after updating to this version, if you
use any pages with __nn__ in their names.
kept in mind during this are a) to reduce load time for common cases like
cgi and post-commit and b) make the code easier to navigate.
This also modularises RCS support to the extent that it should be possible
to drop in a module for some RCS other than svn, add a switch for it, and
it pretty much just work.
High chance I missed an edge case that breaks something, this is only
barely tested at this point.
joey