Ok, I think I undertand what you are looking for.
So basically, you find a bug, create an entry in Mantis, and add the bug number to TL so that you have a trace.
Now unless you are doing some white/grey box testing, your chances of knowing which area of the code cause the problem are slim, unless you have partitioned Mantis very cleanly.
Now the developer takes your bug and resolves it (you are the one who should be closing it, right?).
Where I see an integration required is between Mantis and SVN. Upon check in/commit, the dev. puts the bug number in the commit notes and Mantis changes the state to resolved (As I may have said, I am not very familiar with Mantis).
If they are nice
they give you proper release notes with a candidate, or you find your way through the commit notifications/change log and gather the bug fixes. Regress them, and close them. Et voila.
Some bug tracking system also allow you to gather change sets. Some Fancy one such as Rational UCM/ClearQuest, Accurev will do it automatically.
Doing this kind of integration is in my opinion not trivial (trunk, versus branches, merge points, etc.) so that wouldn't blame the TL dev. not to take this route, because then you have all the other SCM tools. This is a real problem for commercial products as well. I know a few companies that have move their product to the eclipse platform (e.g. Scapa) for that simple reason.
I think that 1.8 is supposed to deliver a plug-in architecture, which will give you the liberty to extend its functionality.
As far as your last question is concerned, well SVN offers a series of hooks (also know as triggers in ClearCase)
In my group, it is for instance located under /directmnt/repository/swrepos/hooks
and there you can find pre-commit, post-commit , email notifications, etc.
Hooks are written Perl.
The TL dev team has been working on a XML-RPC API. I am in the process of testing it internally. What it means is that you need to have their server installed with your TL (I think you need 1.7.3). Because it is XML-RPC you can basically interface TL from pretty much any languages (including Perl).
So bottom line, you process the commit on post or pre commit hook, get the information you need, connect to test-link and update the information at you leisure. Not that complicated to do. The TL team has given some sample clients written in Python and Ruby, I have written my own in Lua.
I use test link in another manner (our automated test runner (will) update the test cases part of a build automatically).
If this is still not what you are looking for, I suggest that you think exactly what you need to be done and file a feature request in Mantis (TestLink).
If the dev team believes that it makes sense for the community, then you might see your wish come true.
Best of luck.
/Laurent
