On Tue, Jun 11, 2013 at 5:44 PM, Erinn Clark <erinn(a)double-helix.org> wrote:
> Another update. Use this one instead:
>
> https://people.torproject.org/~erinn/qa/stable/tor-browser-gnu-linux-x86_64…
> 182a3ff7b5a707cbc84c6919f3c6514e4561c583bf754df0e8c91067f6ab4f73 tor-browser-gnu-linux-x86_64-2.3.25-9-dev-en-US-TEST2.tar.gz
Same shutdown issue with Fedora 18.
On Tue, Jun 11, 2013 at 5:44 PM, Erinn Clark <erinn(a)double-helix.org> wrote:
> Another update. Use this one instead:
>
> https://people.torproject.org/~erinn/qa/stable/tor-browser-gnu-linux-x86_64…
> 182a3ff7b5a707cbc84c6919f3c6514e4561c583bf754df0e8c91067f6ab4f73 tor-browser-gnu-linux-x86_64-2.3.25-9-dev-en-US-TEST2.tar.gz
>
$ uname -a
Linux ubuntu 3.8.0-19-generic #29-Ubuntu SMP Wed Apr 17 18:16:28 UTC
2013 x86_64 x86_64 x86_64 GNU/Linux
Vidalia and Browser starts OK. HTTPS Everywhere is allowing me to
visit Yahoo without SSL/TLS. In fact, entering https://www.yahoo.com
redirects me to http://www.yahoo.com/?s=https. For completeness, it
looks like the IP of the exit node is 222.3.5.198.
Shut down is not clean, and a subsequent `./start-tor-browser` results
in some issues. See the discussion for Debian 6.
Jeff
On Tue, Jun 11, 2013 at 5:44 PM, Erinn Clark <erinn(a)double-helix.org> wrote:
> Another update. Use this one instead:
>
> https://people.torproject.org/~erinn/qa/stable/tor-browser-gnu-linux-x86_64…
> 182a3ff7b5a707cbc84c6919f3c6514e4561c583bf754df0e8c91067f6ab4f73 tor-browser-gnu-linux-x86_64-2.3.25-9-dev-en-US-TEST2.tar.gz
>
$ uname -a
Linux debian-6-x64 2.6.32-5-amd64 #1 SMP Fri May 10 08:43:19 UTC 2013
x86_64 GNU/Linux
Vidalia and Browser started OK. Navigated to a few multimedia sites OK.
Shut down is not clean. After closing Browser and then Control Panel,
components appeared to be lingering in memory:
$ ps -A
...
2625 pts/0 00:00:00 start-tor-brows
2642 pts/0 00:00:10 vidalia
2645 pts/0 00:00:01 tor
2648 ? 00:00:00 sh
2649 ? 00:00:00 run-parts
2653 ? 00:00:00 apt
2676 ? 00:00:00 sleep
2706 pts/1 00:00:00 bash
2719 pts/1 00:00:00 ps
A subsequent `./start-tor-browser` results in some issues. For
example, I was warned about a previous running instance of Vidalia,
and then get prompted for a password (I did not set a password
previously). http://postimg.org/image/hs3u0rcyp/ and
http://postimg.org/image/ft8or2mut/.
Jeff
Finally, we have a 3.0 build that fully matched on two different build
machines for all bundles for the 3.0-alpha-2 series. I'm still waiting
on a third confirmation, but since these bundles contain the 17.0.7-ESR
security release (which is over a week old now), I want to get them out
ASAP.
If anyone is willing to try these and report any major issues, that
would be great.
I will be posting these on the blog as soon as I hear back from Georg
Koppen and get a detached signature from him and Linus Nordberg.
Here's my builds:
https://people.torproject.org/~mikeperry/3.0-alpha-2/
Here's Linus's matching builds:
https://people.torproject.org/~linus/downloads/tbb-3.0alpha2-build5-c0242c2…
Here's the ChangeLog:
https://gitweb.torproject.org/builders/tor-browser-bundle.git/blob/HEAD:/Bu…
Here's how you reproduce your own identical bundles:
https://gitweb.torproject.org/builders/tor-browser-bundle.git/blob/HEAD:/gi…
The tor-browser-bundle.git commit used is c0242c24bed086cc9c545c7bf2d699948792c1e3,
with signed tag tbb-3.0alpha2-build6.
If anyone does this, I'd also love to hear if there are any differences.
--
Mike Perry
Hi All/Mike,
My apologies for spinning up another thread. The original seemed like
it was getting cluttered, and I believe the cause has been found.
In this installation, the Tor bundle was installed in %PROGRAMFILES%.
Specifically, C:\Program Files (x86)\Tor Browser.
If I run "Start Tor Browser.exe" as a regular user, I receive an error
message, "Firefox is already running, but not responding...". See
"tor-firefox-windows-already-running.png" at
http://postimg.org/image/cf8ily0sp/.
When I trace the processes involved (Start Tor, Tor, and Firefox)
using Sysinternal's Process Monitor, it appears Firefox is trying to
write a lock file in %PROGRAMFILES%. The lock file is ...\Tor
Browser\FirefoxPortable\Data\Profile\parent.lock, and it results in an
ACCESS_DENIED. See "tor-firefox-windows-access-denied.png" at
http://postimg.org/image/xlzmb3j1b/.
If I run "Start Tor Browser.exe" as an Administrator, the browser works fine.
When the Tor Browser is launched after installation (as part of the
installation), the browser works fine. I believe this is because the
setup/install program is not dropping privileges after it completes.
In this case, FirefoxPortable should probably be writing its lock file
to a temporary directory or the User's application data directory
(CSIDL_APPDATA or FOLDERID_RoamingAppData,
http://msdn.microsoft.com/en-us/library/windows/desktop/dd378457%28v=vs.85%…).
It should not attempt to write to a read/execute directory.
Jeff
On Fri, Jun 14, 2013 at 10:39 PM, Mike Perry <mikeperry(a)torproject.org> wrote:
> The new TBB 3.0 series is almost ready for its first alpha release!
>
> Here are the major highlights of the 3.0 series:
>
> 1. Usability, usability, usability! We've attempted to solve several
> major usability issues in this series, including:
>
> A. No more Vidalia. The Tor process management is handled by the Tor
> Launcher extension. If you want the Vidalia features, you can
> point an existing Vidalia binary at control port 9151 after Tor
> Browser has launched, and it should still work.
>
> B. The browser now uses a local about:tor homepage instead of
> check.torproject.org. A local verification against the control port
> is still performed, to ensure Tor is working, and a link to
> check.torproject.org is provided from the about:Tor homepage
> for manual verification as well.
>
> C. For Windows users: an NSIS-based extractor now guides you through
> the TBB extraction and ensures the extracted bundle ends up on your
> Desktop, or in a known location chosen by you. Hopefully this
> will mean no more losing track of the extracted bundle files!
>
> 2. The bundles are all under the 25M gmail attachment size limit, so
> direct email and gettor attachments are once again possible.
>
> 3. We now use Gitian to build the bundles. The idea behind Gitian is to
> allow independent people to take our source code and produce exactly
> identical binaries on their own. We're not quite at the point where
> you always get a matching build, but the remaining differences are
> minor, and within a couple more releases we should have it fully
> reproducible. For now, we are posting all of the builds for
> comparison, and you can of course build and compare your own:
> https://gitweb.torproject.org/builders/tor-browser-bundle.git/blob/HEAD:/gi…
>
>
> Please try these out, test them, and give us feedback! The plan is to
> post them on the blog by Monday, unless something goes horribly wrong.
The new TBB 3.0 series is almost ready for its first alpha release!
Here are the major highlights of the 3.0 series:
1. Usability, usability, usability! We've attempted to solve several
major usability issues in this series, including:
A. No more Vidalia. The Tor process management is handled by the Tor
Launcher extension. If you want the Vidalia features, you can
point an existing Vidalia binary at control port 9151 after Tor
Browser has launched, and it should still work.
B. The browser now uses a local about:tor homepage instead of
check.torproject.org. A local verification against the control port
is still performed, to ensure Tor is working, and a link to
check.torproject.org is provided from the about:Tor homepage
for manual verification as well.
C. For Windows users: an NSIS-based extractor now guides you through
the TBB extraction and ensures the extracted bundle ends up on your
Desktop, or in a known location chosen by you. Hopefully this
will mean no more losing track of the extracted bundle files!
2. The bundles are all under the 25M gmail attachment size limit, so
direct email and gettor attachments are once again possible.
3. We now use Gitian to build the bundles. The idea behind Gitian is to
allow independent people to take our source code and produce exactly
identical binaries on their own. We're not quite at the point where
you always get a matching build, but the remaining differences are
minor, and within a couple more releases we should have it fully
reproducible. For now, we are posting all of the builds for
comparison, and you can of course build and compare your own:
https://gitweb.torproject.org/builders/tor-browser-bundle.git/blob/HEAD:/gi…
Please try these out, test them, and give us feedback! The plan is to
post them on the blog by Monday, unless something goes horribly wrong.
https://people.torproject.org/~mikeperry/tbb-3.0alpha1-builds/official/
--
Mike Perry
Hey everyone,
We've been hearing from people for a while that the FF 17.0.x bundles are
unusably slow because we disabled optimization during build time. The reason we
did this is because building on Debian Squeeze resulted in a strange build
crash. I re-built all of the stable 32-bit Linux TBB on Wheezy, with
optimizations re-enabled, and would appreciate it if any of you could test it.
I'm specifically worried about backwards compatibility, so if any of you are
running older Linuxes (like anything CentOS, for example) I'd love to hear if
you have any problems. Likewise, for modern Linuxes, please let me know if the
usability problems have gone.
The package is here: https://people.torproject.org/~erinn/qa/stable/tor-browser-gnu-linux-i686-2…
sha256sum: ed94e886207aa727dd0d7693f9a45beef232dfb3f9df2281ba72961dc6a92e04 tor-browser-gnu-linux-i686-2.3.25-TEST-8-dev-en-US.tar.gz
There's no rapid time limit on this one -- I just need to get some feedback
before upgrading all of the build systems, so being as thorough as possible is
great.
Thanks very much!
Erinn