My testing was complicated by an xtras-sourced message style where the CFBundleName had been changed, but CFbundleIdentifier conflicted with Mockie. However, now that this confusion has been resolved, I can say that The latest fix does appear to resolve the problems with Adium loading message style bundled form the wrong place.
Now I need to test, commit, and push my updated failovers for the old CFBundleIdentifiers. Matthew On Thu, Oct 14, 2010 at 22:33, Evan Schoenberg, M.D. <eva...@dreskin.net> wrote: > Matthew, > > On Oct 12, 2010, at 7:24 PM, Matthew wrote: > >> Not only was this my first download of 3355, it was a completely new >> profile. Sorry I forgot to include the profile creation step. > > I can't reproduce any problems with a first-time upgrade with adium-1.4. A > quick perusal of hg's log shows that I failed to transplant this to the adium > repository, though, which would explain its failure on your side. > > A new profile wouldn't change anything with the current implementation, since > this is using NSUserDefaults to track the status of the upgrade which is > app-specific not profile-dependent. > > Further evaluation: This is currently fragile; it's feasible a user could > launch 1.4, relaunch 1.3, then launch 1.4 again, and then be faced with the > same rather odd-appearing bug. > > I've corrected this fragility and pushed the change to both adium-1.4 and > adium. Please recheck and let me know if you can still reproduce any > problems in this department :) > > Thanks as always for your help! > > Cheers, > Evan > >> >> >> Matthew >> >> On Tue, Oct 12, 2010 at 18:30, Evan Schoenberg, M.D. <eva...@dreskin.net> >> wrote: >>> Had you launched hgr3355 (or any after my changeset) before testing as you >>> wrote below? I wrote the upgrade code to only trigger once. >>> >>> -Evan >>> >>> On Oct 12, 2010, at 6:01 PM, Matthew wrote: >>> >>>> I don't see any change from the old behavior. Here's what I did: >>>> >>>> 1) Launch 1.3.10, and open the messages prefpane so that "Current >>>> Style Path" gets set in WebKit Message Display.plist. >>>> >>>> 2) Quit 1.3.10, launch 1.5hgr3355, and check WebKit Message >>>> Display.plist. "Current Style Path" is still set to >>>> "/Users/mneedham/Applications/Adium >>>> 1.3.10/Adium.app/Contents/Resources/Message >>>> Styles/Stockholm.AdiumMessageStyle", so I'd expect any saved chat >>>> windows will load with the incorrect message style. >>>> >>>> 3) Open the messages prefpane, and verify visually, that the preview >>>> is still for the version of Stockholm in 1.3.10. (In the old version, >>>> there the status message timestamp has a grey background.) I verified >>>> that "Current Style Path" is still set to the message style inside the >>>> 1.3.10 app bundle. >>>> >>>> >>>> Matthew >>>> >>>> >>>> >>>> On Mon, Oct 11, 2010 at 13:23, Matthew <mneed...@ei8ht.us> wrote: >>>>> I'll try to work though my test cases later today or tomorrow. >>>>> >>>>> >>>>> Matthew >>>>> >>>>> On Mon, Oct 11, 2010 at 12:52, Evan Schoenberg, M.D. <eva...@dreskin.net> >>>>> wrote: >>>>>>> On Sep 22, 2010, at 9:07 AM, Matthew wrote: >>>>>> >>>>>>>> Here's what I found: >>>>>>>> >>>>>>>> In the User profile, Webkit Message Display.plist is created when the >>>>>>>> messages prefpane is opened. It contains: >>>>>>>> >>>>>>>> <key>Current Style Path</key> >>>>>>>> >>>>>>>> <string>/Users/mneedham/Applications/Adium-1.3/Adium.app/Contents/Resources/Message >>>>>>>> Styles/Stockholm.AdiumMessageStyle</string> >>>>>>>> >>>>>>>> This key does not change when the user launches b18 or if the user >>>>>>>> launches b18 and opens the messages prefpane. Should it? When I open >>>>>>>> the messages prefpane of 1.4b19, I can see that Stockholm (the default >>>>>>>> style) is using the one bundled with 1.3.10/1.4b18, and *not* the one >>>>>>>> bundled with 1.4b19. I don't know if the Current Style Path key is >>>>>>>> responsible for pointing Adium to the correct message style bundle, or >>>>>>>> if another file is responsible for that (if so I can't find one) I >>>>>>>> think I've been told that Adium writes this value out to two places, >>>>>>>> so maybe I should be looking elsewhere. >>>>>> >>>>>> I've fixed this in http://hg.adium.im/adium-1.4/rev/7eced108f702 >>>>>> >>>>>> -Evan >>>>>> >>>>> >>>>> >>>>> >>>>> -- >>>>> >>>>> Matthew >>>>> >>>> >>>> >>>> >>>> -- >>>> >>>> Matthew >>>> >>> >>> >>> >> >> >> >> -- >> >> Matthew >> > > > -- Matthew
