New Release Candidate: Tor 0.4.3.4-rc

by nickm | April 13, 2020

There's a new alpha release available for download. If you build Tor from source, you can download the source code for 0.4.3.4-rc from the download page on the website. Packages should be available over the coming weeks, with a new alpha Tor Browser release likely later this week.

This is a release candidate: unless we find new significant bugs in it, the stable release for the 0.4.3.x series will be substantially the same as this release.

Tor 0.4.3.4-rc is the first release candidate in its series. It fixes several bugs from earlier versions, including one affecting DoS defenses on bridges using pluggable transports.

Changes in version 0.4.3.4-rc - 2020-04-13

  • Major bugfixes (DoS defenses, bridges, pluggable transport):
    • Fix a bug that was preventing DoS defenses from running on bridges with a pluggable transport. Previously, the DoS subsystem was not given the transport name of the client connection, thus failed to find the GeoIP cache entry for that client address. Fixes bug 33491; bugfix on 0.3.3.2-alpha.
  • Minor feature (sendme, flow control):
    • Default to sending SENDME version 1 cells. (Clients are already sending these, because of a consensus parameter telling them to do so: this change only affects what clients would do if the consensus didn't contain a recommendation.) Closes ticket 33623.

 

  • Minor features (testing):
    • The unit tests now support a "TOR_SKIP_TESTCASES" environment variable to specify a list of space-separated test cases that should not be executed. We will use this to disable certain tests that are failing on Appveyor because of mismatched OpenSSL libraries. Part of ticket 33643.
  • Minor bugfixes (--disable-module-relay):
    • Fix an assertion failure when Tor is built without the relay module, and then invoked with the "User" option. Fixes bug 33668; bugfix on 0.4.3.1-alpha.
  • Minor bugfixes (--disable-module-relay,--disable-module-dirauth):
    • Set some output arguments in the relay and dirauth module stubs, to guard against future stub argument handling bugs like 33668. Fixes bug 33674; bugfix on 0.4.3.1-alpha.
  • Minor bugfixes (build system):
    • Correctly output the enabled module in the configure summary. Before that, the list shown was just plain wrong. Fixes bug 33646; bugfix on 0.4.3.2-alpha.
  • Minor bugfixes (client, IPv6):
    • Stop forcing all non-SocksPorts to prefer IPv6 exit connections. Instead, prefer IPv6 connections by default, but allow users to change their configs using the "NoPreferIPv6" port flag. Fixes bug 33608; bugfix on 0.4.3.1-alpha.
    • Revert PreferIPv6 set by default on the SocksPort because it broke the torsocks use case. Tor doesn't have a way for an application to request the hostname to be resolved for a specific IP version, but torsocks requires that. Up until now, IPv4 was used by default so torsocks is expecting that, and can't handle a possible IPv6 being returned. Fixes bug 33804; bugfix on 0.4.3.1-alpha.
  • Minor bugfixes (key portability):
    • When reading PEM-encoded key data, tolerate CRLF line-endings even if we are not running on Windows. Previously, non-Windows hosts would reject these line-endings in certain positions, making certain key files hard to move from one host to another. Fixes bug 33032; bugfix on 0.3.5.1-alpha.
  • Minor bugfixes (logging):
    • Flush stderr, stdout, and file logs during shutdown, if supported by the OS. This change helps make sure that any final logs are recorded. Fixes bug 33087; bugfix on 0.4.1.6.
    • Stop closing stderr and stdout during shutdown. Closing these file descriptors can hide sanitiser logs. Fixes bug 33087; bugfix on 0.4.1.6.
  • Minor bugfixes (onion services v3):
    • Relax severity of a log message that can appear naturally when decoding onion service descriptors as a relay. Also add some diagnostics to debug any future bugs in that area. Fixes bug 31669; bugfix on 0.3.0.1-alpha.
    • Block a client-side assertion by disallowing the registration of an x25519 client auth key that's all zeroes. Fixes bug 33545; bugfix on 0.4.3.1-alpha. Based on patch from "cypherpunks".
  • Code simplification and refactoring:
    • Disable our coding standards best practices tracker in our git hooks. (0.4.3 branches only.) Closes ticket 33678.
  • Testing:
    • Avoid conflicts between the fake sockets in tor's unit tests, and real file descriptors. Resolves issues running unit tests with GitHub Actions, where the process that embeds or launches the tests has already opened a large number of file descriptors. Fixes bug 33782; bugfix on 0.2.8.1-alpha. Found and fixed by Putta Khunchalee.
  • Testing (CI):
    • In our Appveyor Windows CI, copy required DLLs to test and app directories, before running tor's tests. This ensures that tor.exe and test*.exe use the correct version of each DLL. This fix is not required, but we hope it will avoid DLL search issues in future. Fixes bug 33673; bugfix on 0.3.4.2-alpha.
    • On Appveyor, skip the crypto/openssl_version test, which is failing because of a mismatched library installation. Fix for 33643.

Comments

Please note that the comment area below has been archived.

April 18, 2020

Permalink

What level of technical prowess would you recommend as a minimum for running development versions of TBB?

I really wouldn't consider there to be any "minimum" technical ability to run a development version of the browser bundle. The browser bundle is already the "user-friendly" approach to Tor, so if you want to run the devleopment version, go ahead and do it. It's highly unlikely, though, that you'll notice much of a difference most of the time... but it's even more unlikely that running the dev version of the browser bundle will cause any problems for you. The development version is still a (fairly large) step removed from compiling and running the code from the actual git repository at whatever random commit you happen to clone it at. I've done that dozens of times by the way myself. To do that, you'll want to be proficient in using Linux on the command line, and unless you know at least some C there's really not going to be much of a reason why you would bother to mess with compiling the code straight from the repo anyway. That's interacting with Tor directly, at its core. The browser bundle - even the dev version - is several very significant steps away from that in the direction of user-friendliness. If you want to run the development browser bundle, go ahead and do so, regardless of your technical experience =)