Hello,
So, 7.5a5 is newly broken with sandboxed-tor-browser, unless you're running git master (as of a few mins ago) due to https://trac.torproject.org/projects/tor/ticket/23692
Even with the uncertain future of sandboxed-tor-browser this highlights a few issues that I'm not sure how to address:
* In general, it's not immediately obvious to me how I'm supposed to test for new releases of the browser requiring changes to the sandbox in time for a new sandbox build to get shipped with the new browser release.
* I'm strongly tempted to drop support for the alpha channel, because I run stable builds of the browser, so that's more likely to break since it's a "install it when someone cries about it" thing for me.
Suggestions welcome, in the mean time if people want to use sandboxed-tor-browser with the alpha channel, they need to build git master.
Regards,
On Fri, Sep 29, 2017 at 03:11:54AM +0000, Yawning Angel wrote:
So, 7.5a5 is newly broken with sandboxed-tor-browser, unless you're running git master (as of a few mins ago) due to https://trac.torproject.org/projects/tor/ticket/23692
Thanks for the fix!
- I'm strongly tempted to drop support for the alpha channel, because I run stable builds of the browser, so that's more likely to break since it's a "install it when someone cries about it" thing for me.
Isn't this a nice success case for the alpha channel, where it found a bug and now the bug is fixed?
I mean, I agree it would be smarter to have some sort of QA thing going on before alpha releases too. But if the worry is "nobody tests stuff before release", backing off so people only try the stable isn't going to help the stable stay stable.
--Roger
On Thu, 28 Sep 2017 23:16:38 -0400 Roger Dingledine arma@mit.edu wrote:
- I'm strongly tempted to drop support for the alpha channel,
because I run stable builds of the browser, so that's more likely to break since it's a "install it when someone cries about it" thing for me.
Isn't this a nice success case for the alpha channel, where it found a bug and now the bug is fixed?
Don't know. Depends on if people expect the alpha channel to be usable or not.
At this point there's been what I ordinarily would consider unreasonable periods of time when the alpha release was unusable due to things that I either broke and didn't test for (because I basically exclusively use the stable release), or the browser changing and things breaking that I can't test prior to a release being made.
If it's just a "testing" thing, then this may be fine. Otherwise, delivering something that ends up being broken quite often isn't very responsible in my world view.
I mean, I agree it would be smarter to have some sort of QA thing going on before alpha releases too. But if the worry is "nobody tests stuff before release", backing off so people only try the stable isn't going to help the stable stay stable.
Regardless, until I find something better, git master will keep working with the stable channel, if only because I use it.
"official" releases are an entirely different story, because I have no way of knowing if any given release will work till the release is made presently.
Regards,
On Fri, Sep 29, 2017 at 04:59:46AM +0000, Yawning Angel wrote:
Don't know. Depends on if people expect the alpha channel to be usable or not.
It is true that the right thing for the Tor Browser team to do now is either to take down the alpha sandbox, because it is known to not work at all, or to put up a new updated one.
Leaving the current one in place is not among the reasonable options. :)
--Roger
Roger Dingledine:
On Fri, Sep 29, 2017 at 04:59:46AM +0000, Yawning Angel wrote:
Don't know. Depends on if people expect the alpha channel to be usable or not.
Well, I think assuming issues is okay but we should not leave it broken. Yawning, could you tag a new release? Then we can build a new sandboxed-tor-browser and put that out. Just falling back to an older version won't help.
It is true that the right thing for the Tor Browser team to do now is either to take down the alpha sandbox, because it is known to not work at all, or to put up a new updated one.
Leaving the current one in place is not among the reasonable options. :)
I tend to agree although I think we should avoid just taking it down as it is not totally broken in the sense that choosing any option would result in a non-working browser.
Georg
--Roger
tbb-dev mailing list tbb-dev@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tbb-dev
On Fri, 29 Sep 2017 06:01:00 +0000 Georg Koppen gk@torproject.org wrote:
Roger Dingledine:
On Fri, Sep 29, 2017 at 04:59:46AM +0000, Yawning Angel wrote:
Don't know. Depends on if people expect the alpha channel to be usable or not.
Well, I think assuming issues is okay but we should not leave it broken. Yawning, could you tag a new release? Then we can build a new sandboxed-tor-browser and put that out. Just falling back to an older version won't help.
Done.
It is true that the right thing for the Tor Browser team to do now is either to take down the alpha sandbox, because it is known to not work at all, or to put up a new updated one.
Leaving the current one in place is not among the reasonable options. :)
I tend to agree although I think we should avoid just taking it down as it is not totally broken in the sense that choosing any option would result in a non-working browser.
I mean, the stable channel works, which is what people should be using anyway, and it's not the first time that "oh, alpha is broken, and will be for a while" cropped up, so it could be left as is, though now that I tagged, if someone wants to rebuild they can.
Constructive(?) suggestions for long term solutions:
* I've said previously what I think should happen so that everyone gets the sandbox. Doing so would remove this as a separate component, with a "new and improved" launcher also handling sandboxing duties. The QA situation presumably will be better in this sort of world, because breaking bugs will render the browser non-functional.
* Decouple the sandboxed-tor-browser build/release from the alpha builds, so that the sandbox can be released after it's known to actually work with each new stable/alpha release.
* (All other options involve varying degrees of "keep doing what we're doing now" and "give up on things". Not really a suggestion, more of an indication of what's likely.)
Regards,
Yawning Angel:
On Fri, 29 Sep 2017 06:01:00 +0000 Georg Koppen gk@torproject.org wrote:
Roger Dingledine:
On Fri, Sep 29, 2017 at 04:59:46AM +0000, Yawning Angel wrote:
Don't know. Depends on if people expect the alpha channel to be usable or not.
Well, I think assuming issues is okay but we should not leave it broken. Yawning, could you tag a new release? Then we can build a new sandboxed-tor-browser and put that out. Just falling back to an older version won't help.
Done.
It is true that the right thing for the Tor Browser team to do now is either to take down the alpha sandbox, because it is known to not work at all, or to put up a new updated one.
Leaving the current one in place is not among the reasonable options. :)
I tend to agree although I think we should avoid just taking it down as it is not totally broken in the sense that choosing any option would result in a non-working browser.
I mean, the stable channel works, which is what people should be using anyway, and it's not the first time that "oh, alpha is broken, and will be for a while" cropped up, so it could be left as is, though now that I tagged, if someone wants to rebuild they can.
Okay. The next alpha will be out in less than three weeks. Maybe that's enough and we point the sandbox alpha users (which run an experimental software in an experiemental sandbox) to the stable series?
However, the bug seems to be due to content sandboxing enabled (if I see that right) and my plan was to just ship the stable Linux series with that enabled in about three weeks if nothing explodes. That would leave sandboxed-tor-browser users on the stable series with the same experience with no way upgrading to 0.0.14 before 7.0.7 gets out. We might want to avoid that.
Constructive(?) suggestions for long term solutions:
- I've said previously what I think should happen so that everyone gets the sandbox. Doing so would remove this as a separate component, with a "new and improved" launcher also handling sandboxing duties. The QA situation presumably will be better in this sort of world, because breaking bugs will render the browser non-functional.
I am a fan. We just need to think how we get there.
- Decouple the sandboxed-tor-browser build/release from the alpha builds, so that the sandbox can be released after it's known to actually work with each new stable/alpha release.
We could think about that one, too. Although I would rather have a way to test those releases in sandboxed-tor-browser beforehand than adding complexity to our release process.
Georg
- (All other options involve varying degrees of "keep doing what we're doing now" and "give up on things". Not really a suggestion, more of an indication of what's likely.)
Regards,
tbb-dev mailing list tbb-dev@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tbb-dev
-
Georg Koppen -
Roger Dingledine -
Yawning Angel