Re: ole32: On big endian machines, copy strings to little endian order without mucking with the map they're stored in (try 2)

[Alexandre Julliard]

Juan Lang  writes:
> +#ifdef WORDS_BIGENDIAN
> +/* Returns a byte-swapped copy of str, to be freed with HeapFree(). */
> +static LPWSTR PropertyStorage_StringToLE(LPCWSTR str, size_t len)
> +{
> - LPWSTR leStr = HeapAlloc(GetProcessHeap(), 0, len * sizeof(WCHAR));
> -
> - if (leStr)
> - {
> -
>    memcpy(leStr, str, len * sizeof(WCHAR));
> -
>    PropertyStorage_ByteSwapString(leStr, len);
> - }
> - return leStr;
>
> +}
> +#else
> +#define PropertyStorage_StringToLE(s, l) (LPWSTR)(s)
> +#endif

You are still casting away const...

ole32: On big endian machines, copy strings to little endian order without mucking with the map they're stored in (try 2)

[Juan Lang]

> You are still casting away const...

Indeed I am, but now the warning is harmless:  the const pointer is
only read from, never written to.  Making a const pointer in the
little endian case would make the entire block rather messier, since
having a non-const pointer is necessary in the big endian case:  it's
written to and freed.

If you'd rather I introduce #ifdefs in the block itself, let me know.
--Juan

ole32: On big endian machines, copy strings to little endian order without mucking with the map they're stored in (try 2)

[Alexandre Julliard]

Juan Lang  writes:
> > You are still casting away const...
>
> Indeed I am, but now the warning is harmless:  the const pointer is
> only read from, never written to.  Making a const pointer in the
> little endian case would make the entire block rather messier, since
> having a non-const pointer is necessary in the big endian case:  it's
> written to and freed.
>
> If you'd rather I introduce #ifdefs in the block itself, let me know.

No, but you can just wrap the whole thing in a function.