<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.
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
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!
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>
All reviewed patches have been landed. Closing bug.
<rdar://problem/51814778>
(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.
*** Bug 198646 has been marked as a duplicate of this bug. ***