#216 gcc compiler warnings

[Jan Nijtmans]

When compiling the current rc with gcc 4.6 , I get two compiler warmings:

./generic/itcl_objects.c: In function ‘Itcl_CreateObject’:
./generic/itcl_objects.c:195:25: warning: passing argument 5 of ‘tclStubsPtr->tcl_TraceVar2’ from incompatible pointer type [enabled by default]
./generic/itcl_objects.c:195:25: note: expected ‘char * (*)(void *, struct Tcl_Interp *, char *, char *, int)’ but argument is of type ‘char * (*)(void *, struct Tcl_Interp *, const char *, const char *, int)’
./generic/itcl_parse.c: In function ‘Itcl_ParseInit’:
./generic/itcl_parse.c:147:9: warning: passing argument 3 of ‘tclIntStubsPtr->tcl_SetNamespaceResolvers’ from incompatible pointer type [enabled by default]
./generic/itcl_parse.c:147:9: note: expected ‘int (*)(struct Tcl_Interp *, char *, struct Tcl_Namespace *, int, struct Tcl_Var_ **)’ but argument is of type ‘int (*)(struct Tcl_Interp *, const char *, struct Tcl_Namespace *, int, struct Tcl_Var_ **)’

The reason is that the signatures of the functions Itcl_ClassVarResolver
and Itcl_ParseVarResolver don't match the signature expected by
Tcl_TraceVar2 resp Tcl_SetNamespaceResolvers

Here is the fix for that.

[Jan Nijtmans]

proposed fix

[Jan Nijtmans]

alternative (additional) fix

[Jan Nijtmans]

Alternative (additional) fix.

In stead if fixing those 2 signatures, we could also
stop compiling Itcl with -DUSE_NON_CONST.
patch (itcl2.patch) attached.

Either of those two patches fix the compiler
warning. itcl.patch makes everything compile
file with -DUSE_NON_CONST, itcl2.patch
makes everything compile file without
-DUSE_NON_CONST.

You can even apply both patches, if
you like. Then everything will compile
file no matter what USE_NON_CONST
is set to. Just apply what you think best.

(I think I would apply itcl2.patch only:
no-one will ever compile Itcl
with -DUSE_NON_CONST
if the Makefile does not have
it already ....)

[Don Porter]

Does Itcl 3.4 even build against Tcl 8.3 ?
Is it important that it continue to do so?

[Don Porter]

The branch of development now called "itcl-3-branch"
in fossil has not been buildable against Tcl 8.3 since
August of 2007. To me that's strong evidence there is
no one left who needs it to build against Tcl 8.3.

Assuming that is correct, the obvious right answer is to
drop all the CONST shenanigans, since their only purpose
is to maintain 8.3 support. Just update every interface to
the CONST conventions of the 8.4 interface.

[Don Porter]

See the drop-pre-Tcl-8-4 branch for all the
proposed changes to Itcl 3.4 .

[Jan Nijtmans]

@dgp: I don't mind about building againsy Tcl 8.3 (never tried it), so I can only agree with all
those proposed additional changes. itcl2.patch was a first step in that direction, but
changing all CONST84's to simply CONST would be the next step. +1

[Don Porter]

pre-8.4 Tcl support dropped from Itcl 3.4.1