message.txt

Mark,

Thanks for the e-mail and the comments. I've made a couple of changes which will be in the next version, see comments below.

On the changes for compiling under VC++:
It would be good to have a list of the changes, and the compiler variable which identifies VC++ so that I could add something like
#ifdef MS_VC // or whatever it uses
    wsprintf(etc);
#else
    wvsprintf(etc);
#endif

I'll put the next set of sources up this weekend, but if I get these in then it would help you recompile the next version (which mainly contains a source change from word segm, dword offs to lptr loc and a small class for the lptr loc with meaningful (in the context) comparisons,etc and a lot of simplification of code.

Debugging is difficult in Borland (debugger won't run win32 apps), so it's useful to have the feedback,

Thanks
Cronos

Cronos
In porting beta10 to VC++ I believe have found a couple of memory bugs, both are in
names::addname, the problems may occur elsewhere aswell.

1) at the end of the function there is a test
   if (!strlen(newname->name)) return;

this leaves newname and newname->name as unreferenced memory i.e. a memory leak
// Yes, you're quite right.
// -added delete newname->name; delete newname; here.

2) the code
  addto((listitem)newname);
  dsm.discomment(segm,offs,dsmnameloc,(byte *)newname->name);

puts name on two lists, at the end of the program the delete functions for the lists both try to
release the memory, this could lead to win95 GPFs, debugging with VC++ certainly flagged problems on exit.
// yes, this seems to be right too.
// -have added 'and !=dsmnameloc in the dsm deletion function too, before deleting tptr item.

A oddity in your code, which may be a bug (I haven't had time to check) is in fileloader::readsysfile,
you use offsets into fbuff of 0, 3 & 4. Given that fbuff appears to be a byte array these offsets don't seem right.
// fbuff is cast to a (word *) - careful bracketing !

A few extra points on the port

a) you use wvsprintf, VC++ complains since it cannot safely cast the va_list arg, the solution I used was to
use wsprintf instead and remove the & prefixes to the arguments.
// yes, i've had some trouble with wvsprintf in Borland, but it seems to work in Borland now.

b) VC++ complained about dasm.rc because there was no windows.h include file also there is an extra
comma on the save_database line.
// Borland doesn't complain.

Hope the input is useful

regards
Mark Ogden

PS. Have you considered using Mapping to load the files rather than reading into a buffer?
// Please explain further !