diff options
| -rw-r--r-- | ChangeLog | 9 | ||||
| -rw-r--r-- | generic/tclObj.c | 24 |
2 files changed, 28 insertions, 5 deletions
@@ -1,3 +1,12 @@ +2009-12-10 Andreas Kupries <andreask@activestate.com> + + * generic/tclObj.c (TclContinuationsEnter): [Bug 2895323]: Updated + comments to describe when the function can be entered for the same + Tcl_Obj* multiple times. This is a continuation of the 2009-11-10 + entry where a memory leak was plugged, but where not sure if that + was just a band-aid to paper over some other error. It isn't, this + is a legal situation. + 2009-12-10 Miguel Sofer <msofer@users.sf.net> * generic/tclBasic.c: Reducing the # of moving parts for diff --git a/generic/tclObj.c b/generic/tclObj.c index 5913dd1..667fd90 100644 --- a/generic/tclObj.c +++ b/generic/tclObj.c @@ -13,7 +13,7 @@ * See the file "license.terms" for information on usage and redistribution of * this file, and for a DISCLAIMER OF ALL WARRANTIES. * - * RCS: @(#) $Id: tclObj.c,v 1.164 2009/11/10 17:57:39 andreas_kupries Exp $ + * RCS: @(#) $Id: tclObj.c,v 1.165 2009/12/10 19:13:26 andreas_kupries Exp $ */ #include "tclInt.h" @@ -584,10 +584,24 @@ TclContinuationsEnter(Tcl_Obj* objPtr, if (!newEntry) { /* - * Somehow we're entering ContLineLoc data for the same value (objPtr) - * more than one time. Not sure whether that's expected, or a sign of - * trouble, but at a minimum, we should take care not to leak the old - * entry. + * We're entering ContLineLoc data for the same value more than one + * time. Taking care not to leak the old entry. + * + * This can happen when literals in a proc body are shared. See for + * example test info-30.19 where the action (code) for all branches of + * the switch command is identical, mapping them all to the same + * literal. An interesting result of this is that the number and + * locations (offset) of invisible continuation lines in the literal + * are the same for all occurences. + * + * Note that while reusing the existing entry is possible it requires + * the same actions as for a new entry because we have to copy the + * incoming num/loc data even so. Because we are called from + * TclContinuationsEnterDerived for this case, which modified the + * stored locations (Rebased to the proper relative offset). Just + * returning the stored entry would rebase them a second time, or + * more, hosing the data. It is easier to simply replace, as we are + * doing. */ ckfree((char *) Tcl_GetHashValue(hPtr)); |