svn:r17229

svn:r17228

Add some notes on Unnamed, w lines, and p lines to dir-spec.txt, since they are all in the code now.

svn:r17227

to for many years. i think.


svn:r17224

svn:r17217

svn:r17216

svn:r17215

svn:r17214

Aaaaand, do not reset the download count just because we have a certificate that is listed in a consensus.

svn:r17212

svn:r17211

When we get a duplicated certificate, treat it as a failure and increment the download count. Do not claim to be downloading certificates that we merely want.

svn:r17209

svn:r17208

svn:r17207

it's obsolete. which causes us to inform the user every time, even
though the user can't do anything about it other than get confused.

now it's an info-level log by default.


svn:r17206

the bottom of switch_id


svn:r17205

idempotent. so now we remember if we've succeeded, and if so we
don't even try.


svn:r17204

svn:r17203

svn:r17202

Developers should usually configure with --enable-gcc-warnings, and should regularly make check-spaces.  Also, int fn() does not mean the same in C as it does in C++ or Java.

svn:r17201

Patch from Jacob Appelbaum and me to make User option more robust, properly set supplementary groups, deprecated the Group option, and log more information on credential switching

svn:r17200

svn:r17199

svn:r17198

svn:r17195

svn:r17194

svn:r17193

svn:r17192

svn:r17191

svn:r17190

(a - b) / 2 != a - (b / 2); this is the reason why tunneled directory requests and client-side introduction circuits don't time out after the intended 30 seconds (task #847). Bugfix on r17106. Found by miner.

svn:r17189

svn:r17188

Add a new memcmpstart to use instead of strcmpstart when the thing we are comparing is not nul-terminated.

svn:r17187

svn:r17186

svn:r17185

chosen exit node in mind already. otherwise we get tricked into
trying to build a new circuit that will handle them.


svn:r17184

we were complaining about no support for our one-hop streams,
when in fact choose_good_exit_server_general() has no business
caring about one-hop streams. patch from miner.


svn:r17181

svn:r17180

svn:r17179

Implement the 0x20-hack to make DNS poisoning harder against us, especially when resolving large names.  Add a cfg option to disable it, since apparently 3/10 of a percent of servers get it wrong.

svn:r17171

Work better with tools that resist DNS poisoning by using the 0x20 hack: make DNSPort replies perserve case.

svn:r17170

svn:r17169