Bug 147243 - jsc-tailcall: Repatching tail calls as jump should depend on the opcode, not the JS CallLinkInfo
Summary: jsc-tailcall: Repatching tail calls as jump should depend on the opcode, not ...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Basile Clement
URL:
Keywords:
Depends on:
Blocks: 146477
  Show dependency tree / graph
 
Reported: 2015-07-23 16:46 PDT by Basile Clement
Modified: 2015-07-24 11:59 PDT (History)
2 users (show)

See Also:

Attachments
Patch (9.29 KB, patch)
2015-07-23 16:50 PDT, Basile Clement 		no flags Details | Formatted Diff | Diff
Patch (9.71 KB, patch)
2015-07-24 11:26 PDT, Basile Clement 		msaboff: review+
Details | Formatted Diff | Diff
Show Obsolete (1) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Basile Clement 2015-07-23 16:46:03 PDT 
Patch forthcoming. The apparent source of failures on ARM devices for https://bugs.webkit.org/show_bug.cgi?id=146484 looks to be that we are patching NearCalls as jump when the CallLinkInfo tells us we are a tail call.

However even for a tail call, the slow path will still have an actual call (and not jump) opcode: we should be patching based on that information instead.
Comment 1 Basile Clement 2015-07-23 16:50:49 PDT 
Created attachment 257408 [details]
Patch
Comment 2 Michael Saboff 2015-07-24 11:03:54 PDT 
Comment on attachment 257408 [details]
Patch

Looks fine except could you change the "bool tail" to an enum.
Comment 3 Basile Clement 2015-07-24 11:26:56 PDT 
Created attachment 257461 [details]
Patch

Patch with enum
Comment 4 Michael Saboff 2015-07-24 11:32:22 PDT 
Comment on attachment 257461 [details]
Patch

r=me
Comment 5 Basile Clement 2015-07-24 11:59:15 PDT 
Landed in https://trac.webkit.org/changeset/187354.