I swear we've seen this 0x80000022 error a few different places related to using the wrong/outdated Xcode utils.. I can't find the exact bug I'm thinking about, but #2813 (closed) is similar.
After carefully re-reading the emails you sent me and this bug report, my question is: which bundle is the user actually using? The email says x86_64 which definitely won't work on 10.5.8 -- that is only built with the 10.6 SDK, and the crash logs are exactly what I was seeing when the i386 bundle wasn't being built with the 10.5 backwards compatibility options. Can you verify which bundle is being used, and ask theuser (re-)try the i386 to double check?
There's a new Vidalia in this release, so that might be the culprit. Can you try replacing TorBrowser_en-US.app/Contents/MacOS/Vidalia.app with the Vidalia.app from 0.2.2.35-5 and let me know if that one runs?
There's a new Vidalia in this release, so that might be the culprit. Can you try replacing TorBrowser_en-US.app/Contents/MacOS/Vidalia.app with the Vidalia.app from 0.2.2.35-5 and let me know if that one runs?
I can confirm that proposed process works (after replacing Vidalia.app, Vidalia starts, Firefox 10.0.2 starts and succesfully connects to the Tor network).
Tried TBB 2.2.35-7 32bit (with Vidalia.app from 0.2.2.35-5) on Mac OS X 10.5.8.
Hi all, this is my first time posting to a bug tracker, so I apologize in advance if I'm contributing incorrectly/incompletely.
I just tried the above with the new 0.2.3.12 alpha browser bundle for Mac. That bundle did not open for me at all; when I opened the package contents and ran Tor Browser Bundle, it failed showing error -10810.
I read this ticket and tried replacing Vidalia from Browser Bundle 2.2.35-5 for i386. Now Vidalia opens and connects to Tor, but Firefox never starts. If I open try to open Firefox form its package contents, the terminal shows this:
rickv$ /Users/rickv/Applications/TorBrowser_en-US.app/Contents/MacOS/Firefox.app/Contents/MacOS/firefox ; exit;
2012-03-11 14:15:58.131 firefox[51151:10b] *** _NSAutoreleaseNoPool(): Object 0x17609cc0 of class NSCFDictionary autoreleased with no pool in place - just leaking
Stack: (0x93be6f4f 0x93af3432 0x2a0f16c)
logout
It is obvious that Vidalia is the cause of the problem. If is latest Vidalia compiled different way, why not to compile it again - the same way as it was compiled for TBB 2.2.35-5?
Since 2.2.35-5 for OS X, running Vidalia via command line ends with following:
I am using above mentioned workaround, replacing distributed Vidalia.app folder with old one (last working from 2.2.35-5), however - it doesn't seems to be the right way to solve the issue (using outdated Vidalia in the new bundle)...
And another question: is this problem persisting for OS X 10.6 and 10.7 users or is it only 10.5.8 anomally?
It is obvious that Vidalia is the cause of the problem. If is latest Vidalia compiled different way, why not to compile it again - the same way as it was compiled for TBB 2.2.35-5?
Because it's not obvious what the problem is. It looks like the old build options together with new Vidalia code produce the problematic behaviour.
And another question: is this problem persisting for OS X 10.6 and 10.7 users or is it only 10.5.8 anomally?
It is obvious that Vidalia is the cause of the problem. If is latest Vidalia compiled different way, why not to compile it again - the same way as it was compiled for TBB 2.2.35-5?
Because it's not obvious what the problem is. It looks like the old build options together with new Vidalia code produce the problematic behaviour.
And another question: is this problem persisting for OS X 10.6 and 10.7 users or is it only 10.5.8 anomally?
Yes, only 10.5 is affected.
In this case, it is obvious that the application was incorrectly built on OS X 10.6 (or newer) machine for a 10.5 machine. A developer can fix this by considering three things:
Using the correct compiler parameters:
gcc-4.2 -mmacosx-version-min=10.5 -isysroot /Developer/SDKs/MacOSX10.5.sdk...
Using the correct linker settings (setting environment variable before link command). This is required, so that the OS X 10.6 linker will not use the loader command 'LC_DYLD_INFO_ONLY' (=0x80000022), because OS X 10.5 does not understand it.
It is obvious that Vidalia is the cause of the problem. If is latest Vidalia compiled different way, why not to compile it again - the same way as it was compiled for TBB 2.2.35-5?
Because it's not obvious what the problem is. It looks like the old build options together with new Vidalia code produce the problematic behaviour.
I can't comment on this. The build maintainer claims (and I have no reason to not believe her) that the last working bundle was also built on 10.6.
And another question: is this problem persisting for OS X 10.6 and 10.7 users or is it only 10.5.8 anomally?
Yes, only 10.5 is affected.
In this case, it is obvious that the application was incorrectly built on OS X 10.6 (or newer) machine for a 10.5 machine. A developer can fix this by considering three things:
Using the correct compiler parameters:
gcc-4.2 -mmacosx-version-min=10.5 -isysroot /Developer/SDKs/MacOSX10.5.sdk...
Using the correct linker settings (setting environment variable before link command). This is required, so that the OS X 10.6 linker will not use the loader command 'LC_DYLD_INFO_ONLY' (=0x80000022), because OS X 10.5 does not understand it.
Thanks for the lesson. Maybe instead of putting random bugs in needs_review without actually providing patches, you could've checked out the current way things are getting built? We do set backwards compatibility options for 10.5.
Thanks for the lesson. Maybe instead of putting random bugs in needs_review without actually providing patches, you could've checked out the current way things are getting built?
I did, se below (typos in your build-files etc).
We do set backwards compatibility options for 10.5.
Thanks for your input. I have provided complete guide on what's wrong with the latest binaries, how to check if it's correctly compiled backwards-compatible for 10.5.x on a 10.6 machine and how to link it the right way. I've tested that complete process above myself - using last git source (0.2.17) from your repository and succesfully compiled an executable, working on 10.5.8.
What more should I do?
Since today, I'm going to compile backwards-compatible vidalia binary myself when the maintainer can't.
BTW - check out /pkg/osx/build-bundle.txt, line 36 (https://gitweb.torproject.org/vidalia.git/blob_plain/HEAD:/pkg/osx/build-bundle.txt) clearly states (in regards to bundled QT compilation):
MACOSX_DEPLOYENT_TARGET=10.4.
See the typo? May wrongly set env. variable relate to wrongly linked binaries and copy+pasting?! Questions. Nevermind.