Summary: | [chromium] Test WebSolidColorLayer in WebLayerTest.Client unit test | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | James Robinson <jamesr> | ||||
Component: | New Bugs | Assignee: | James Robinson <jamesr> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | abarth, enne, fishd, webkit.review.bot | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
Attachments: |
|
Description
James Robinson
2012-09-04 22:44:03 PDT
Created attachment 162160 [details]
Patch
The test value here is marginal, but this is important for the component build in a rather convoluted way. When use_libcc_for_compositor=1, the implementations of the Web*Layer::create() family of static functions are in a static_library webkit/compositor_bindings/compositor_bindings.gyp:webkit_compositor_bindings instead of in a bunch of .cpp files in the webkit sources list. Because of this, these symbols are linked into a separate lib that is then linked into webkit.dll. If any symbols from this .lib are completely unreferenced, they're dropped from webkit.dll even if they have WEBKIT_EXPORT set on them. The fixes are to explicitly pull these in by either a linker flag (like -force_load, /INCLUDE, or /DEF) or to make sure they're referenced in webkit.dll. Since we link the unit tests into webkit.dll, that appears to do the trick. Comment on attachment 162160 [details]
Patch
That seems a bit hacky, but ok.
More than a bit! The test coverage is marginally useful on its own, though. Comment on attachment 162160 [details] Patch Clearing flags on attachment: 162160 Committed r127678: <http://trac.webkit.org/changeset/127678> All reviewed patches have been landed. Closing bug. |