The optional Lookup.framework soft-link hits a release assertion when it's missing. This regressed r240494 when I changed to use the newer soft-linking macros, but the newer soft-linking macros originally broke in r238955 when optional frameworks starting hitting release assertions. <rdar://problem/47924449>
Created attachment 361767 [details] Patch v1
Created attachment 361771 [details] Patch v2
Created attachment 361790 [details] Patch v3 The charm.
(In reply to David Kilzer (:ddkilzer) from comment #3) > Created attachment 361790 [details] > Patch v3 > > The charm. Looks like the iOS bots don't agree with this old adage.
(In reply to Eric Carlson from comment #4) > (In reply to David Kilzer (:ddkilzer) from comment #3) > > Created attachment 361790 [details] > > Patch v3 > > > > The charm. > > Looks like the iOS bots don't agree with this old adage. Doh. Should have rebuilt again locally after adding DataDetectorsCoreSoftLink.mm to the WebCore target.
Created attachment 361801 [details] Patch v4 Fix iOS builds.
Comment on attachment 361801 [details] Patch v4 View in context: https://bugs.webkit.org/attachment.cgi?id=361801&action=review > Source/WebCore/ChangeLog:21 > + * platform/cocoa/DataDetectorsCoreSoftLink.mm: > + - Switch from using SOFT_LINK_PRIVATE_FRAMEWORK_OPTIONAL() to > + SOFT_LINK_PRIVATE_FRAMEWORK_FOR_SOURCE() when linking > + DataDetectorsCore.framework. None of the other macros assume > + this framework is optional, and it was likely made optional > + originally because the framework was new to iOS and thus > + didn't exist on older versions. Also, the optional behavior for frameworks in these new-style soft-linking macros really depends on how they're called from the other Class, Pointer, Constant, Function, etc. soft-linking macros in the same source file.
Comment on attachment 361801 [details] Patch v4 Clearing flags on attachment: 361801 Committed r241309: <https://trac.webkit.org/changeset/241309>
All reviewed patches have been landed. Closing bug.