At 06:38 AM 2/17/00 -0800, Martin Buchholz wrote: could be Of course I have a feeling I've broken your C++ compile once again. andy -------------------------------------------------------------- Dr Andy Piper Senior Consultant Architect, BEA Systems Ltd

kkm> I compiled XEmacs with a C++ compiler yesterday night, and it pointed kkm> me to this bug (":warning: Empty control statement. Is ';' in intended place?") kkm> Same error is also present somewhere in DND code. Patch follows. kkm> I think that C++ compile is a very good idea. I've found several bugs this way. Because autoconf and lib-src are not C++-clean, you need to compile with a C compiler and a C++ compiler. With configure, use the undocumented env XEMACS_CC=/msc/c++ configure ... With Windows native, you are breaking new ground. I would guess normal make with C compiler; (cd lwlib; make clean;) (cd src; make clean;) make CC=/msc/c++ would work. kkm> A question. I have fixed all errors and warnings which C++ compiler gave, kkm> except for conversion errors between char* and unsigned char*. This is kkm> an error in C++: kkm> unsigned char str[20]; kkm> strset (str, 'z', 19); kkm> I do not really want to use casts at such places - they certainly do not kkm> improve on readability of the code. Any luck avoiding these? I know how you feel. But I think you should fix those, because: - this is an ERROR, and we'd like to do things like USE the C++-compiled code, especially for `make check' and performance hacking. - this is generally a WARNING under C, so these should be fixed anyways; not just for C++. - I think it was a mistake to have typedef'ed Extbyte, Bufbyte, etc. as `unsigned char', despite the added safety of consistent signedness of chars. So there are many casts for (char *) to (Bufbyte *), etc... _If_ we are serious about avoiding warnings and avoiding casts, then we should create a collection of string handling function versions of the functions taking (char *) args. INLINE size_t Extlen (Extbyte *p) { return strlen ((char *) p); } INLINE size_t Buflen (Bufbyte *p) { return strlen ((char *) p); } INLINE size_t ustrlen (unsigned char *p) { return strlen ((char *) p); } ... etc... ... This is my kind of thing, and is something the project could agree to do. But it will be a major global code change with no direct user benefit... In fact, actual users would suffer from slightly longer compile times... Anyways, I recommend, for now, doing the distasteful strset ((unsigned char *) str, 'z', 19); Martin P.S. I don't know what `strset' is, but I could guess and suggest str[19] = 'z';