518615 - [10.6] Using Snow Leopard and Growl, first download results in all windows losing focus

Status: RESOLVED FIXED

(Camino Graveyard :: Downloading, defect)

Description

[David Munch]

User-Agent:       Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en; rv:1.9.0.15pre) Gecko/2009091516 Camino/2.0b4 (like Firefox/3.0.15pre)
Build Identifier: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en; rv:1.9.0.15pre) Gecko/2009091516 Camino/2.0b4 (like Firefox/3.0.15pre)

The first download during a new session, results in all Camino windows loosing focus. This only happens with Mac OS X 10.6 and if the user has Growl installed.



Reproducible: Always

Steps to Reproduce:
1. Start Camino.
2. Download something.
3. Download windows pops up, and download starts.
4. All windows looses focus.


Expected Results:  
Download window should retain focus.

Apparently this is a known problem to the Growl developers. Listening in on an IRC conversation (As good as I could, not being a programmer), they think it is an OS bug. And they won't fix it until Growl 1.3 is out, which is going to take quite a while.

Also, they told me:
"find shawn and tell him he'll need to pay attention to the 1.3 release cycle" & "and for now I think you guys are just going to have the problem" & "and could work around it in your code if you want to" & "we'll accept patches only in bsd 3 clause if that comes up".

Comment 1

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

...Which is presumably the bug described in the URL field.

If no fix is going to be forthcoming, we'll almost certainly have to relnote this for 2.0, which is always fun to do for new features :P

Comment 2

[Chris Forsythe]

And of course you only got half the story. :)

Growl 1.3 drops the current communication protocol in favor of our new protocol, gntp. It's mostly done, some testing and cryptographic features need to be implemented. So if we can't fix this in 1.2, it'll be gone in 1.3.

That said, try beta 5 of our 1.2 release. Well, specifically, the changes to the framework that attempt to address this. Outside of mozilla products, we only see this sporadically with GrowlSafari as of yet, so it's in your benefit to take a look:

http://growl.info/beta.html

Also there are a few tickets lying around on multiple mozilla products for this (firefox for instance has a long one) that would be worth looking at.  Bug 515396 and  Bug 510740 most likely will provide some context here.

Is this a duplicate ticket? I would think you guys would have had a ticket on this by now.

Comment 3

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #2)
> http://growl.info/beta.html

No source tarball for the beta? (We land third-party source code in our repo and build, rather than landing binary components.)

> Also there are a few tickets lying around on multiple mozilla products for this
> (firefox for instance has a long one) that would be worth looking at.  Bug
> 515396 and  Bug 510740 most likely will provide some context here.

Bug 510740 was originally filed by a Camino user (before we knew it was Growl that was the problem and instead thought it was some Gecko problem), and once we knew it was Growl, Rudy helped Ilya debug our integration (without a separate bug).

> Is this a duplicate ticket? I would think you guys would have had a ticket on
> this by now.

We'd already fixed the bug 510740 case, but, surprisingly, David is the only person who has reported this new variant so far.  However, we'd been following bug 515396 so we knew it was very likely Camino would have an equivalent bug.

Comment 4

[Chris Forsythe]

(In reply to comment #3)
> (In reply to comment #2)
> > http://growl.info/beta.html
> 
> No source tarball for the beta? (We land third-party source code in our repo
> and build, rather than landing binary components.)
> 

Check it out from mercurial. We only tarball up final releases.

> > Also there are a few tickets lying around on multiple mozilla products for this
> > (firefox for instance has a long one) that would be worth looking at.  Bug
> > 515396 and  Bug 510740 most likely will provide some context here.
> 
> Bug 510740 was originally filed by a Camino user (before we knew it was Growl
> that was the problem and instead thought it was some Gecko problem), and once
> we knew it was Growl, Rudy helped Ilya debug our integration (without a
> separate bug).
> 
> > Is this a duplicate ticket? I would think you guys would have had a ticket on
> > this by now.
> 
> We'd already fixed the bug 510740 case, but, surprisingly, David is the only
> person who has reported this new variant so far.  However, we'd been following
> bug 515396 so we knew it was very likely Camino would have an equivalent bug.


Ah, ok, makes sense. Just didn't want duplicates floating around.

Comment 5

[David Munch]

Growl 1.2b5 does not fix it for me, unfortunately.

Comment 6

[Stuart Morgan]

Did you replace the framework inside of the Camino bundle?

Comment 7

[David Munch]

Ah, no I did not, I only installed Growl 1.2b5. I'm going to need install directions, or a test build for that. :)

Comment 8

[Stuart Morgan]

Putting on the 2.0? list so we remember to try out the framework update; we wouldn't hold for this though.

Comment 9

[Stuart Morgan]

Er, I guess that's what the milestone field is for. Sorry for bugspam.

Comment 10

[Stuart Morgan]

(In reply to comment #7)

1) Download http://growl.googlecode.com/files/Growl-1.2-SDK.dmg
2) Show Package Contents on Camino.app
3) Open Contents/Frameworks
4) Trash Growl.framework
5) Replace it with the Growl.framework from the Frameworks folder on the Growl disk image.

Comment 11

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Test build with Growl 1.2+: http://www.ardisson.org/smokey/moz/Camino-2.1a1pre-2009-10-01-21-Growl1.2.dmg

David, does this fix the bug for you?

Comment 12

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Update to Growl 1.2+, Camino part + README

On the forum, it sounds like ozean is not seeing the bug in the test build, although it's hard to tell whether he noticed the bug to begin with, so some additional confirmation from David would be great.

That said, here's the deal:

* Upgrade to Growl 1.2 plus Rudy's fix for http://code.google.com/p/growl/issues/detail?id=26
  - This adds a few new file compared to what we have checked in and also 
    removes a couple of files.
  - In addition, Rudy's fix allows us to remove a couple of headers previously 
    checked in on which the framework source no longer depends.
* Fix a few additional details in the project, as outlined in README:
  - NEW - Converted to Xcode 3.0 format to make it more compatible with Xcode 
    2.5 toolchain on 10.4.
  - Commented out the SDK declarations to support mozconfig-based 
    configurations.
  - NEW - Changed the default build configuration to Release.
  - NEW - Supplied valid archs for building with toolchains on Mac OS X 10.4 
    and 10.5.
  - NEW - Set the MACOSX_DEPLOYMENT_TARGET to 10.4 in all configurations.
  - Disabled the incompatible save-repository setting to prevent bustage when 
    building on 10.4.
* Fix up the Camino project, Makefile, and xcconfig to account for the new 
  new-style config names in Growl.xcodeproj

I'll attach 3 patches:

* camino/ changes, plus camino/growl/README, for review
* Growl.xcodeproj vs upstream 1.2+ Growl.xcodeproj, for sanity-checking
* All camino/growl/ changes, for reference (and diffscrape-aided checkin)

I've built with these changes with the Xcode 3 toolchain on 10.5, and in Xcode 2.5 on 10.5 (which hopefully will translate to success with the Xcode 2.5 toolchain on 10.4).

Comment 13

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Update to Growl 1.2+, Growl project changes vs upstream project

Growl.xcodeproject vs. its upstream version, for sanity-checking

Comment 14

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Update to Growl 1.2+, Growl part

All changes within mozilla/camino/growl, for reference

Comment 15

[Stuart Morgan]

Comment on attachment 404709 [details] [diff] [review]
Update to Growl 1.2+, Camino part + README

No problems here, sr=smorgan

Comment 16

[Stuart Morgan]

Comment on attachment 404710 [details] [diff] [review]
Update to Growl 1.2+, Growl project changes vs upstream project

> 				"MACOSX_DEPLOYMENT_TARGET[arch=i386]" = 10.4;
> 				"MACOSX_DEPLOYMENT_TARGET[arch=ppc]" = 10.4;

My experience from Sparkle is that this notation makes Xcode on 10.4 very angry, so we should axe these.

> 				"OTHER_CFLAGS[arch=i386]" = "-fsingle-precision-constant";
> 				"OTHER_CFLAGS[arch=ppc]" = "-fsingle-precision-constant";

Anywhere we have things like this (that aren't 64-bit only) we should replace with a non-arch version, so we don't change the build settings that end up used. Skimming the other diff I see OTHER_CFLAGS, and a couple of GCC_ENABLE_OBJC_GC where there is a value, but it's wrong (i.e., different than the ppc/i386 value).

Comment 17

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Update to Growl 1.2+, Growl project changes vs upstream project, v2

I searched for all instances of "[arch" and either commented out the entire block (all that 64-bit fun) or commented out the block and replaced it with a generic equivalent where one was not already present.

Comment 18

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #17)
> I searched for all instances of "[arch" and either commented out the entire
> block (all that 64-bit fun) or commented out the block and replaced it with a
> generic equivalent where one was not already present.

…And added a line about that to the README, too.

Comment 19

[David Munch]

Sorry about the late response, totally forgot about this bug! Bug me on IRC the next time.. :)

Anway, the test-build posted by Ardisson did fix the problem for me! \o/

Comment 20

[Stuart Morgan]

Comment on attachment 404770 [details] [diff] [review]
Update to Growl 1.2+, Growl project changes vs upstream project, v2

> 				GCC_ENABLE_OBJC_GC = supported;
>-				"GCC_ENABLE_OBJC_GC[arch=i386]" = unsupported;
>-				"GCC_ENABLE_OBJC_GC[arch=ppc]" = unsupported;
>+				/* "GCC_ENABLE_OBJC_GC[arch=i386]" = unsupported;
>+				"GCC_ENABLE_OBJC_GC[arch=ppc]" = unsupported; */

No change to unsupported for the base value? I guess this is with with-installer version though, so it doesn't matter.

sr=smorgan

Comment 21

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Landed on cvs trunk and CAMINO_2_0_BRANCH.

Thanks, David, for confirming the test build worked :-)