I believe I have found a bug.
If you goto http://www.etherpunk.com/comment/reply/180 (possibly NSFW) it allows you to view the posting (while you don't have permission to actually post, it still allows the page to get displayed).
I found this out by using awstats on my box and found that a hidden page was getting hit fairly common that I really don't want getting shown (well, it's on the web, I know... but... I'd rather have more controlled access).

Does this belong in the comment section for not obeying TAC?

Can anyone confirm this on their site?

Kenny

Support from Acquia helps fund testing for Drupal Acquia logo

Comments

pyromanfo’s picture

That's definitely something you need to take up with the comment module guys. It's not just taxonomy access control either, it's the core node_access hooks in Drupal. If they'll just check that before displaying a node for reply, that'd fix it no problem.

moshe weitzman’s picture

Project: Taxonomy Access Control » Drupal core
Component: Code » comment.module
Priority: Normal » Critical

filed under comment.module ... note that my big comment patch gets rid of this page entirely (consolidates under comment/edit) so it might make sense to apply my patch instead of fixing this.

chx’s picture

Assigned: Unassigned » chx
FileSize
1.36 KB

moshe , http://drupal.org/node/18656 this does not seem to affect the permissions of the comment/reply path.

I think the approach I have taken is blatantly simple: literally check for access.

chx’s picture

chx’s picture

chx’s picture

chx’s picture

Anonymous’s picture

The discovery of this patch makes me wonder, whether we shouldn't centralize the access controll a bit more.
If a core module can show this kind of bug, contrib modules will almost certainly. I propose to do a node_access() check inside node_load.
Gerhard

Anonymous’s picture

I also don't think that the patch is working.
node_access(arg(2)) should probably be node_access('view', $node) and the node needs loading before.
Gerhard

chx’s picture

OK, patch corrected.

I think adding node_access to node_load would break havoc 'cos there are some routines (even in core) which do not check whether the node_load was successful or not. This is a whole another topic, please make another issue.

What we need, IMHO is to patch this quickly...

chx’s picture

I also forgot to check whether node_load was successful or not...

chx’s picture

Anonymous’s picture

The patch looks good now. I agree that centralizing the node_access checking should be another task. This patch is needed now and should also be backported to the 4.5 branch.
Gerhard

chx’s picture

Dries commited.

killes@www.drop.org’s picture

Here's a 4.5 patch. Untested.

Anonymous’s picture