On some Windows machines the browser window is white during start-up
I noticed that while testing bundles on my Windows system that after subsequent starts a confusing empty browser window is showing up during the Tor Launcher dialog being visible. That never happened before Tor Browser 9, though.
Today a confused user showed up on #tor reporting what I believe is actually this issue. Attached is a screenshot.
Activity
On some Windows machines But should be on all. https://wiki.mozilla.org/Platform/Integration/InjectEject/Launcher_Process/
FWIW: From memory, I saw this on almost every non en-US pack (never on en-US). I set up a full suite of 9.07a (32 bundles at the time, non English) and they're still all sitting there for when I get around to updating them for more tests.
I just reproduced at will on the first random test - russian. Windows 7. Is there a command line I can run to debug and output anything for you.
My initial thought was this smells a bit like the sandboxing race condition bug fixed in #26381 (moved) (during testing it also showed a white content page) but it's not clear from the screenshot that this is the same symptom.
-
aaaaand .... this is the ticket I was actually after
^^ tl;dr: AFAICT it's something to do with language builds. I can reproduce at will. Once again, if you want, give me instructions on how to debug for you. FYI: flipping
browser.startup.blankWindowwill mask the symptom -
^^ my bad. I should have updated
I updated my language test suite (and added the few new ones since). All 34 non_en-US builds are Windows 64-bit, with spoof english on. There was a smattering of different versions (as I usually just pick a random language for an occasional test: and sometimes manually update), but most were
9.07a(ouch). I think the latest was about9.5a4. I used the in-app updating mechanism.Before updating, every single one (I can't pinky swear: maybe one didn't: I took no notes) would show the unpainted window before the tor connection dialog. After being prompted to restart, on the restart the problem did not occur. Also tested, just starting them (i.e not restarts), and the problem does not happen anymore.
As for en-US: I have never had that problem on any build: and I run 32 and 64 bit stables and alphas.
I noticed a lot of translation fixups (still a few strings missing in some languages, and some strings in all languages): maybe something in there inadvertently fixed this? Like that broken
:vendorStringThing?So, unless you want to test 32-bit, and/or non-one else can replicate anymore, then I think this can be closed?
