This is non-optimal default behaviour, if repo_dir isn't just a bunch of repos this could potentially slow things down a lot.
Basically Syntax::Highlight::Kate is slow and trying to prettify anything relative big s really slow. So just don't even try. Sad face.
Previously blame would get stuck in an infinite loop when processing an empty file. Conceivably this a bug in git-blame as file content should begin with a tab, even if it's empty. Also adding a commit to repo1 involves fiddling lots tests so I won't be doing that again. Thanks to autarch for flagging this up in the first place!
Previously the icon only looked sane on a white background, anything else and you could see ugly white splotches. Now the background is fully transparent. Also cleaned up unused favicon.ico paths.
Since the owner name is retrieved from the system user table and thus can end up being anything.
Looks like the latest pod2text now includes the links from the POD in the resulting text, which is nice.
Will also integrates Test::CPAN::Changes when it is AuthorRequires friendly.
The .psgi was written by franky (Francoise Dehinbo) and gist-ed by Ranguard (Leo Lapworth). I've tweaked it to play nice with checkouts and "make install".
The last of Tom's work is now integrated with all URLs that can will respond with JSON. Unfortuantely browsers (I'm looking at you Chrome) will cache per URL rather than URL + Content-Type so if one requests JSON for /repo/sha1/commit then navigates to said commit the browser will served the cached JSON rather than HTML. As such JSON is also exposed with the .json extension allowing JS to talk JSON without messing up certain browsers' (i.e Chrome's) caching.
Another first attempt once more based on Tom's earlier work. Much still to be built on.
This is very much based on the work t0m did in his json branch (which was merged into swann to become swann-with-json). The nicely broken up Git objects created more replicated code than I like but it seems to work which is enough for now. Next up is the rest of the objects and then onto a shiny new REST interface.
Fix other bits that look wonky after the background drop shadow was removed.
Logic isn't quite the same for both, which lead to paging fail, so I've split them in two.
My way of understanding things - make them simply enough so they fit in my brain.
This was dropped when FromDirectoryRecursive was added. But seeing as how it has been the default behaviour for some time now search_recursively is enabled by default in gitalist_local.conf.
The "/" in the name of the repo was not escaped. Use uri_for_action to generate the link to the fragments URI. That implies deleting the "around" uri_for_action in Gitalist.pm. That seems to be OK because tests still pass. Also, didn't find any invocation to uri_for_action with a /fragments path.
Fix model_collectionofrepos.t tests
Make sure that two repos with the same name are retrievable, and are different.
Now the names of the repos that DirectoryRecursive returns are relative paths. Also add two repos with the same name, in different directories to the test suite.
Refactor subclasses to have the _get_repo_from_name as the mandatory method.
Thanks to Kent Fredric for pointing this out in RT#66525.
The current shadows and rounded corners don't quite line up and match so we just use plain ol' CSS instead.
Also try to render commit messages faithfully.
Use this capability in Git::CollectionOfRepositories::FromDirectoryRecursive
Also tidied up existing gitalist.conf and other bits and pieces.
Seemed a shame to have a feature that was only enabled given a particular directory configuration.
This makes text files that have UTF8 bits in them render normally. Certainly this is not ideal but it shall suffice.