602460 - cfx/mozrunner can't launch Firefox nightlies on OS X 10.5

Status: VERIFIED FIXED

(Testing :: Mozbase, defect)

Description

[Drew Willcoxon :adw]

Firefox OS X nightlies recently switched to a new i386 + x86_64 universal binary (bug 571367), and running firefox-bin directly, either from the command line or from cfx, results in this on OS X 10.5:

  dyld: unknown required load command 0x80000022
  Trace/BPT trap

You can run

  arch -i386 firefox-bin

to easily work around the problem, but I don't know how to tell cfx/mozrunner to do that without modifying the code.  Bug 598507 is a similar situation to Jetpack -- they needed to fix Talos scripts there -- and they ended up using this arch command.

Nickolay filed bug 602049, so CC'ing him in case he has thoughts on how to fix this for cfx/mozrunner.

Comment 1

[Drew Willcoxon :adw]

Attachment: patch

How's this?  Seems to work OK on my machine for both old-style universal binaries (Firefox 3.6) and new-style (nightly).

Comment 2

[Atul Varma [:atul]]

Comment on attachment 481466 [details] [diff] [review]
patch

Cool, looks good to me. Thanks!

My only added suggestion is that I like to add a comment of "see bug XXX for more information", since it gives readers a pointer to find out how we arrived at the particular solution we implemented, and a venue for continuing the conversation at a later time by adding another comment to the bug, reopening it, etc.

Er, wait, so I just noticed this is a modification to mozrunner. Are we officially "forking" mozrunner at this point in the jetpack-sdk, or do we want to commit this change and then push it upstream separately, or push it upstream and then refresh our mirror on the jetpack-sdk side?

Comment 3

[Drew Willcoxon :adw]

(In reply to comment #2)
> Er, wait, so I just noticed this is a modification to mozrunner. Are we
> officially "forking" mozrunner at this point in the jetpack-sdk, or do we want
> to commit this change and then push it upstream separately, or push it upstream
> and then refresh our mirror on the jetpack-sdk side?

I don't know.  I checked the MozRunner bugzilla component but didn't see any bugs about this.  Sure seems like it should be fixed in MozRunner though.  Let's move this bug to MozRunner and see what happens!

Comment 4

[Drew Willcoxon :adw]

Clint, what should we do here?  Can you guys take this patch?

Comment 6

[Heather Arthur [:harth]]

http://github.com/mozautomation/mozmill/commit/997d8e3ef20fd98bcad1f11762987e887dcf67fb

Comment 7

[Atul Varma [:atul]]

It looks like this patch's code to determine whether the host platform is 32-bit returns false positives, so I've created bug 604539 to address that.

Comment 8

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Works fine on 10.5 with recent nightly builds. Marking as verified fixed.