Bug 147475

Summary: jsc-tailcall: Add enums for type-safety
Product: WebKit Reporter: Basile Clement <basile_clement>
Component: JavaScriptCoreAssignee: Basile Clement <basile_clement>
Status: RESOLVED FIXED    
Severity: Normal CC: msaboff
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: Unspecified   
OS: Unspecified   
Bug Depends on:    
Bug Blocks: 146477    
Attachments:
Description Flags
Patch
msaboff: review-
Patch
none
Patch
none
Patch msaboff: review+

Basile Clement
Reported 2015-07-30 16:57:34 PDT
We should have: - An enum to specify whether we want to keep or trash the caller frame instead of using reinterpret_cast<void*>(0) and reinterpret_cast<void*>(1) - An enum to indicate whether we are performing a regular call, a tail call, or a constructor call.
Attachments
Patch (29.20 KB, patch)
2015-07-30 17:03 PDT, Basile Clement 		msaboff: review-
DetailsFormatted DiffDiff
Patch (29.10 KB, patch)
2015-07-30 17:08 PDT, Basile Clement 		no flags
DetailsFormatted DiffDiff
Patch (31.91 KB, patch)
2015-07-30 17:34 PDT, Basile Clement 		no flags
DetailsFormatted DiffDiff
Patch (32.06 KB, patch)
2015-07-30 17:38 PDT, Basile Clement 		msaboff: review+
DetailsFormatted DiffDiff
Show Obsolete (3) View All Add attachment proposed patch, testcase, etc.
Basile Clement
Comment 1 2015-07-30 17:03:44 PDT
Created attachment 257881 [details] Patch
Basile Clement
Comment 2 2015-07-30 17:04:53 PDT
Building to run the tests but this should solve the build failures from the merge from trunk.
Basile Clement
Comment 3 2015-07-30 17:08:31 PDT
Created attachment 257882 [details] Patch Now it builds
Michael Saboff
Comment 4 2015-07-30 17:28:21 PDT
Comment on attachment 257881 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=257881&action=review r- You need to add CallMode.{cpp,h} & DeferredSourceDump.{cpp,h} to Windows and cmake builds. > Source/JavaScriptCore/bytecode/CallMode.h:31 > +enum class CallMode { Tail = 0, Regular, Construct }; Minor nit - Wouldn't it make more logical sense to order these Regular, TailCail, Construct, is there special meaning to TailCall being 0? > Source/JavaScriptCore/bytecode/CallMode.h:33 > +enum FrameAction { KeepTheFrame = 0, TrashTheFrame }; Doesn't ReuseTheFrame sound better?
Basile Clement
Comment 5 2015-07-30 17:30:47 PDT
(In reply to comment #4) > Comment on attachment 257881 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=257881&action=review > > r- > You need to add CallMode.{cpp,h} & DeferredSourceDump.{cpp,h} to Windows and > cmake builds. Oh oops, I knew there was a reason I kept this locally instead of submitting a patch earlier... (Note that there is nothing to do w/ DeferredSourceDump, it was just reordered when I did "Sort by name" in XCode). > > > Source/JavaScriptCore/bytecode/CallMode.h:31 > > +enum class CallMode { Tail = 0, Regular, Construct }; > > Minor nit - Wouldn't it make more logical sense to order these Regular, > TailCail, Construct, is there special meaning to TailCall being 0? It would make sense. Tail is 0 because CallMode and FrameAction were a single enum initially, changing this. > > > Source/JavaScriptCore/bytecode/CallMode.h:33 > > +enum FrameAction { KeepTheFrame = 0, TrashTheFrame }; > > Doesn't ReuseTheFrame sound better? ReuseTheFrame is much nicer, changing this.
Basile Clement
Comment 6 2015-07-30 17:34:59 PDT
Created attachment 257885 [details] Patch
Basile Clement
Comment 7 2015-07-30 17:38:24 PDT
Created attachment 257887 [details] Patch
Michael Saboff
Comment 8 2015-07-30 17:40:30 PDT
Comment on attachment 257887 [details] Patch r=me
Basile Clement
Comment 9 2015-07-30 17:44:01 PDT
Landed in r187629 <http://trac.webkit.org/changeset/187629>
Note You need to log in before you can comment on or make changes to this bug.