Christian Boos wrote: > The href() semantic change could also have other unattended consequence, > for example for plugins.
True. The change affects constructs like: href() + path in the case where href() returned the empty string, that is, when the Trac instance is at the root of a site. The patch fixes the only two instances of this in 0.11-stable. As another data point, a scan through all the plugins on trac-hacks shows that the following plugins are affected: authformplugin (for 0.9) *authopenidplugin autoqueryplugin menusplugin *openidplugin personalreportsplugin secsessionplugin (for 0.10) The following could be affected: dbauthplugin (for 0.9, 0.10) revtreeplugin The plugins marked with * also contain bad uses of href() that are fixed by the new behavior, namely they previously set the "path" parameter of the "trac_auth" cookie to the empty string instead of "/". > So maybe we should considerer reverting r7930 > on the branch, keep it on trunk and mention it as an API change for 0.12 > (TracDev/ApiChanges/0.12). Then either fix #8059 differently or consider > it fixed for 0.12 only. At this point, I'm not sure anymore. The [7930] change makes the usage of href more consistent, so I would suggest keeping it in the long run. And if you think the transition should be made in 0.12, I'm ok with it. So if there are no objections, I'll revert [7930] on 0.11-stable and fix #8059 differently tonight. Then we could quickly discuss if the change to Href makes sense or not, and either document it for 0.12 or revert it on trunk as well. -- Remy
signature.asc
Description: OpenPGP digital signature
