Bug 198904 – REGRESSION (r242686): Paths in XPC services’ LC_DYLD_ENVIRONMENT are incorrect in built products directory

• Component: WebKit2
• Status: RESOLVED FIXED

<https://trac.webkit.org/r242686> changed the location of the XPC service executables relative to the frameworks and libraries in the built product directories, but didn’t update the load commands that set DYLD_FRAMEWORK_PATH and DYLD_LIBRARY_PATH. Consequently, the services don’t load the built frameworks, libraries and shim unless something sets the __XPC_DYLD_… environment variables to point to the right place.

Patch forthcoming.

Created attachment 372223 [details]
Use correct relative paths in load commands when not installing

<https://trac.webkit.org/r245965> should also be reverted after this.

[Darin Adler]

Comment on attachment 372223 [details]
Use correct relative paths in load commands when not installing

View in context: https://bugs.webkit.org/attachment.cgi?id=372223&action=review

Is there a good way to test this?

> Source/WebKit/ChangeLog:16
> +          frameworkâs XPCServices directory predicated on whether this is an install build, rathher

rather unusual spelling (extra h typo)

(In reply to Darin Adler from comment #2)
> Comment on attachment 372223 [details]
> Use correct relative paths in load commands when not installing
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=372223&action=review
> 
> Is there a good way to test this?

Download the build product from EWS, and execute the XPC service binary in a shell (it will crash, but there should not be errors about symbols missing). Doing this just now I see that the patch is wrong, so I am going to rework it.

> 
> > Source/WebKit/ChangeLog:16
> > +          frameworkâs XPCServices directory predicated on whether this is an install build, rathher
> 
> rather unusual spelling (extra h typo)

Created attachment 372225 [details]
Use correct relative paths in load commands when not installing

[Alex Christensen]

Comment on attachment 372225 [details]
Use correct relative paths in load commands when not installing

DEPLOYMENT_LOCATION doesn't seem to be a yes/no kind of variable name to me.

(In reply to Alex Christensen from comment #5)
> Comment on attachment 372225 [details]
> Use correct relative paths in load commands when not installing
> 
> DEPLOYMENT_LOCATION doesn't seem to be a yes/no kind of variable name to me.

It’s documented as having Value Type of Boolean, and offers a Yes/No pop-up menu in Xcode’s Build Settings Editor.

Thanks for the review!

[WebKit Commit Bot]

Comment on attachment 372225 [details]
Use correct relative paths in load commands when not installing

Clearing flags on attachment: 372225

Committed r246506: <https://trac.webkit.org/changeset/246506>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/51814778>

[Darin Adler]

(In reply to mitz from comment #1)
> <https://trac.webkit.org/r245965> should also be reverted after this.

Are you going to do that? Should Alex do that? Is someone else volunteering to do it?

(In reply to Darin Adler from comment #10)
> (In reply to mitz from comment #1)
> > <https://trac.webkit.org/r245965> should also be reverted after this.
> 
> Are you going to do that? Should Alex do that? Is someone else volunteering
> to do it?

I intend to do it.

(In reply to mitz from comment #11)
> (In reply to Darin Adler from comment #10)
> > (In reply to mitz from comment #1)
> > > <https://trac.webkit.org/r245965> should also be reverted after this.
> > 
> > Are you going to do that? Should Alex do that? Is someone else volunteering
> > to do it?
> 
> I intend to do it.

Just confirmed with an r246507 archive from nightly.webkit.org that it’s safe to removed.

[Darin Adler]

*** Bug 198646 has been marked as a duplicate of this bug. ***