… I'd checked for that :S Possibly only partially..)
…en the function to determine the backend gets called; since it was being called from the history buffer as well, the algorithm looking at the current working directory etc was seeing a different value from when it was first invoked. I think the long term solution is to pass the backend through to any function that uses it, rather than working it out internally -- this is more referentially-transparent anyway, we just need to make sure that we don't thread too much state through and flood the api. The other fix involved setting a buffer-local variable with the value of the backend, and using that later.
…current buffer is taken into consideration when the backend is determined.
…uch shorter now too.
…et), and seems mostly complete. Should probably tidy up a bit with customisation groups, etc, and of course a readme.
… theory, allows the user to add their own backends by customising tv-backend-list and tv-determine-backend-function)
…just a cons cell) containing the point location and buffer for each visited tag.
…ags (admittedly, the gtags backend hasn't been tested yet!). Two issues, apart from hackishness: I think it should use something like memq indexing rather than symbol interning, and secondly the gtags backend will pollute the visited buffers with markers that weren't there before, which I hadn't thought about. Might be better to rewrite things in terms of buffer/position rather than markers.
…ch backend to use (ie, gtags, etc. Warning: backends not implemented yet :))
… with optional context for etags only. Still to come of course, support for gtags, and manipulation of the tags structure. Also, autodetection of which we should be using.