[PATCH] Invalid links when trac is on accessed from top-most path

[HumanInternals]

The links in that ticket were considered as being spam, so I added \notspam\ to prevent that from happening. Ignore it.

My Trac installation is accessed via ​http://trac.kelso/. When I add a menu, like that:

query = enabled
query.href = /query
query.label = Custom Query
query.order = 1
query.parent = tickets

It's being rendered as <a \notspam\ href="//query">Custom Query</a>. The // at the beginning of the href causes Chrome and Firefox (and possibly others) to think its a link to "http://query/".

When I remove the / prefix from the href in the configuration, the link becomes <a \notspam\ href="query">Custom Query</a>, which doesn't work when I'm on a URL like 'http://trac.kelso/ticket/123', as the link is pointing to 'http:trac.kelso/ticket/query'.

It works fine when Trac is installed under a sub directory, such as http://trac.kelso/humaninternals/, as the link is rendered as <a \notspam\ href="/humaninternals/query">Custom Query</a>. The issue is when its rendered with two slashes at the beginning.

[HumanInternals]

Attaching a fix that worked for me - using req.href() with an argument, instead of concating it with the href.

[anonymous]

The last patch doesn't work properly, as the href is encoded (so characters like '?' won't work). It's not possible to pass as an argument to req.href(), so I'm uploading a new patch right-strips the base href from '/' characters (href=value.startswith('/') and req.href().rstrip('/')+value or value)

[HumanInternals]

a working patch, as described in comment:2 (was written by me, but I forgot to login)

[Ryan J Ollos]

Duplicate of #7399.

[Ryan J Ollos]

(In [11038]) Fixes #7399, Refs #7680, #8877, #8953: (0.1.1) Links were invalid when Trac was accessed from the top-most path.

[Ryan J Ollos]

(In [11039]) Fixes #7399, Refs #7680, #8877, #8953: (0.1.1) [11038] was the wrong solution. This fix should work better.

[Ryan J Ollos]

Replying to anonymous:

The last patch doesn't work properly, as the href is encoded (so characters like '?' won't work). It's not possible to pass as an argument to req.href(),

I tried some examples using the patch described in ​comment:10:ticket:7399, and it seems to work fine to pass the value to req.href.

Here is one of the examples I tried:

/timeline?from=01%2F17%2F2012&daysback=30&changeset=on&milestone=on&wiki=on&update=Update

[Ryan J Ollos]

(In [11154]) Fixes #7399, Refs #7680, #8877, #8953: Allow Href to take care of path joining. Tested in 0.11 and 0.13dev-r10909.

[Ryan J Ollos]

(In [11155]) Fixes #7399, Refs #7680, #8877, #8953: Removed egg and pydev files accidentally committed in [11154].

[Ryan J Ollos]

(In [11156]) Fixes #7399, Refs #7680, #8877, #8953: Removed egg and pydev files accidentally committed in [11154].

[Ryan J Ollos]

Replying to rjollos:

I tried some examples using the patch described in ​comment:10:ticket:7399, and it seems to work fine to pass the value to req.href.

Here is one of the examples I tried:

/timeline?from=01%2F17%2F2012&daysback=30&changeset=on&milestone=on&wiki=on&update=Update

My statement here is incorrect. This does not work. I'll be trying to resolve it over in #7399.

[Ryan J Ollos]

(In [11166]) Fixes #7399, Refs #7680, #8877, #8953: Yet another attempt to fix invalid links when req.href = '/'. If the href for a menu item specifies a relative path, the trailing forward slashes are stripped from req.href() so that the two cases are handled: 1) req.href() is '/' and 2) req.href() is '/somepath'.

[Ryan J Ollos]

I'm working a similar issue in #9788 for the BookmarkPlugin, and considering that href = req.path_info + value might be equivalent to and cleaner than the code fragment (req.href().rstrip('/') + value).