Tcl Source Code

View Ticket
Login
Ticket UUID: 1706840
Title: LITERAL_ON_HEAP flag unused
Type: Bug Created on: 2007-04-24 18:42:51
Submitter: dgp Assigned to: msofer
Subsystem: 47. Bytecode Compiler Severity:
Priority: 8 Last modified: 2009-12-11 10:17:38
Status: Open Closed by:
Resolution: None Closed on:
Version: obsolete: 8.6a1
Description:
TclRegisterLiteral() accepts
a LITERAL_ON_HEAP flag that
no caller uses.  Will any caller
use it?  Or is this added complexity
we simply do not need?
User Comments:
msofer added on 2007-04-25 02:07:55:
Logged In: YES 
user_id=148712
Originator: NO

I can't see where we'd want to use in the core; also grepped tbcload and tclcompiler without finding a match. Who else?

hobbs added on 2007-11-28 01:23:04:
Logged In: YES 
user_id=72656
Originator: NO

You have to look back in time.  This used to be used to know that you had the data in Tcl-controlled heap memory already, and would save another alloc.  By doing without ... did we lose efficiency?

msofer added on 2008-06-28 20:59:03:
Logged In: YES 
user_id=148712
Originator: NO

I do not recall this ever being used, and can't see how it could in the core: if we are compiling, the literal name is part of the source string and we need a fresh copy. In other cases (e.g. varReform variable names), we have a full Tcl_Obj to use as literal and no copy is needed.

If no use case appears, I will remove the extra complexity before beta.