SHGetSpecialFolderPath in order to find application data folder.

Apparently, until IE 4 (!?) came out, nobody realized that programmers
might like to get paths as strings.  Clearly, a fancy pseudo-OO list
of "identifiers" is a far more convenient way to deal with these
things.  And while we're being OO, why return object that you can free
with free()?  Instead, let's make the user get a handle to an abstract
allocation object, and ask it to free the fancy list, and then ask it
to release itself.  Won't that be fun and convenient?

Navigating ancient Win32 APIs is like bikini-waxing creatures from HP
Lovecraft: to do a good job you must understand what's going on... but
the understanding itself can blast your sanity.


svn:r2480

Remove need for dirservers file: now, we note trusted dirservers in configuration options, and only need to remember addr:port and key digest for each one.


svn:r2479

svn:r2470

plus make router_nickname_is_approved() more obviously unused


svn:r2469

svn:r2468

svn:r2467

svn:r2466

svn:r2465

svn:r2464

svn:r2463

Make directories get generated properly again; debug key lookup.  Also, it is hard to find entries in a table when you never put them there.


svn:r2462

svn:r2461

plus unshadow some variables


svn:r2460

svn:r2459

Check fingerprint list before generating running-routers lists; also add as-yet-uncalled function to update an OR connection nickname on the fly


svn:r2458

svn:r2457

fix a potential confusion in fetch_from_buf_http()
make all our int config options non-negative
better bounds checking on options that are ports


svn:r2456

and clean some deadweight from util.h


svn:r2455

svn:r2454

svn:r2453

Make a bounds-checking replacement for strtol with slightly easier error-checking in the common case


svn:r2452

svn:r2451

svn:r2450

svn:r2449

svn:r2447

svn:r2446

svn:r2445

svn:r2444

Add an implementation of trusted_dir_server logic to switch towards simply remembering an addr/port/keyid for each trusted dir server


svn:r2443

svn:r2442

svn:r2441

svn:r2440

svn:r2439

sometimes circuit_get_open_circ_or_launch() can return 0 but not
return a circuit, e.g. because too many circuits have failed recently
so the new one didn't launch. we need to tolerate that.


svn:r2438

We don't think this is exploitable, but best to fix it anyway.


svn:r2437

svn:r2436

what went wrong


svn:r2435

svn:r2434

svn:r2433

svn:r2432