Bug 33902 – Opcode.h use const void* for Opcode cause error #1211 for RVCT compiler

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Lyon Chen]

In commit r52231, Opcode is changed from void* to const void* when COMPUTED_GOTO is used, this caused a compiling error #1211 for RVCT 4.0 compilers.

Suggest to change to:

--- a/JavaScriptCore/bytecode/Opcode.h
+++ b/JavaScriptCore/bytecode/Opcode.h
@@ -196,7 +196,11 @@
     #undef VERIFY_OPCODE_ID
 
 #if HAVE(COMPUTED_GOTO)
+#if COMPILER(RVCT)
+    typedef void* Opcode;
+#else
     typedef const void* Opcode;
+#endif
 #else
     typedef OpcodeID Opcode;
 #endif

[Lyon Chen]

Created attachment 47035 [details]
Patch for bug 33902

[Darin Adler]

Comment on attachment 47035 [details]
Patch for bug 33902

Two problems here:

    1) We don't accept patches without change log entries.

    2) This is probably not something that should be ifdef'd based on compiler. If void* works for the other compilers, then the patch should just switch from const void* to void*.

[Lyon Chen]

Darin:

In commit r52231, [e-mail] mentioned the change from void* to const void* is to avoid generating bloated codes in interpreter.cpp, so maybe it's better #ifdef this change?

And maybe RVCT compiler can someday updated to support const void* for computed goto, instead of void*?

Hope to hear from [e-mail] on this matter.

[Geoffrey Garen]

> In commit r52231, [e-mail] mentioned the change from void* to const
> void* is to avoid generating bloated codes in interpreter.cpp, so maybe it's
> better #ifdef this change?

I'm not sure if that specific part of the change is required. I guess you'd have to test to find out.

[Lyon Chen]

Created attachment 47128 [details]
New patch with changelog.

[Geoffrey Garen]

Did you verify that this change doesn't change code generation for non-RVCT
compilers?

[Lyon Chen]

No, I didn't, Garen. I thought your comment means this part is not essential to your optimization, and based on Darin's comments, I simply removed the #ifdef for RVCT.

Sorry I don't have the time to verify it for other platforms for now, should I add these #ifdef back?

[Yong Li]

+George Staikos

[Lyon Chen]

Created attachment 47131 [details]
New patch with RVCT flag.

Add a new patch with code change limited to RVCT compiler.

[Darin Adler]

Comment on attachment 47128 [details]
New patch with changelog.

I’ll let Geoff handle reviewing duties on this bug. I’m not sure I know what the issues are and I know he does.

[Maciej Stachowiak]

Comment on attachment 47131 [details]
New patch with RVCT flag.

Seems fine to have this scoped to COMPILER(RCVT)

[WebKit Commit Bot]

Comment on attachment 47131 [details]
New patch with RVCT flag.

Rejecting patch 47131 from commit-queue.

[e-mail] does not have committer permissions according to http://trac.webkit.org/browser/trunk/WebKitTools/Scripts/webkitpy/committers.py.

- If you do not have committer rights please read http://webkit.org/coding/contributing.html for instructions on how to use bugzilla flags.

- If you have committer rights please correct the error in WebKitTools/Scripts/webkitpy/committers.py by adding yourself to the file (no review needed).  Due to bug 30084 the commit-queue will require a restart after your change.  Please contact [e-mail] to request a commit-queue restart.  After restart the commit-queue will correctly respect your committer rights.

[WebKit Commit Bot]

Comment on attachment 47131 [details]
New patch with RVCT flag.

Clearing flags on attachment: 47131

Committed r53891: <http://trac.webkit.org/changeset/53891>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.