>
> So then I think your point, translated to implementation requirements,
> would be to:
>
> 1) Use a 256 byte buffer for -multisel => 0 (why waste memory)
> 2) Use a 4000 byte buffer for -multisel => 1 (allow more files to be
> selected normally, even if the user hasn't caught on to doing memory
> management)
> 3) For simplicity, and smaller numbers, one could just multiply
> -multisel by 4000 with a minimum of 256 bytes for the zero case, and
> achieve all these goals. And a 4000 byte granularity on allocation of
> the buffer is probably not a big deal for people that need big buffers.
>
> I've made these adjustments. I'll hold off committing them until next
> week, in case there are other comments.
>
I think it's not a problem to upgrade default stack buffer to 4096 by
default.
Stack allocation is generally same speed for 256 and 4096.
And only dynamicly allocate a buffer when desired size isn't enough.
Something like that :
void
GetOpenFileName(...)
PPCODE:
...
char filename[4096];
...
filename[0] = '\0';
ofn.lpstrFile = filename;
ofn.nMaxFile = 4096;
...
} else if(strcmp(option, "-multisel") == 0) {
next_i = i + 1;
SwitchBit(ofn.Flags, OFN_ALLOWMULTISELECT,
SvIV(ST(next_i)));
if ( ofn.nMaxFile < (MAX_PATH * SvIV(ST(next_i))) ) {
ofn.nMaxFile = MAX_PATH * SvIV(ST(next_i));
ofn.lpstrFile = (char *) safemalloc(ofn.nMaxFile);
ofn.lpstrFile[0] = '\0';
}
}
...
if(retval) {
if (ofn.Flags & OFN_ALLOWMULTISELECT) {
...
if (ofn.lpstrFile != filename)
safefree(ofn.lpstrFile);
XSRETURN(i);
...
Laurent.
