Re: Lisp_String to LPWSTR

[Eugen Anghel]

2009/10/3 Eli Zaretskii <address@hidden>:
>> From: Eugen Anghel <address@hidden>
>> Date: Sat, 3 Oct 2009 00:59:39 +0300
>>
>> What would be the best way to convert from a Lisp_String* to an UTF-16
>> encoded string?
>>
>> I've tried something like
>> code_convert_string_norecord(mystring, Qutf_16, 0)
>> but that doesn't seem to do what I want.
>>
>> I'm trying to work on this (from etc/TODO):
>> ** Switch the Windows port to using Unicode keyboard input (maybe).
>> and I need to convert strings before passing them to W-variant winapi 
>> functions.
>
> Like this (untested):
>
>   code_convert_string_norecord (SDATA (mystring), Qutf_16, 1);
>
> Note the two differences:
>
>  . To get a C string from a Lisp string, use SDATA (btw, why do you
>    manipulate Lisp strings here?)

code_convert_string_norecord is defined as
Lisp_Object code_convert_string_norecord P_ ((Lisp_Object, Lisp_Object, int));
so I don't think the SDATA is needed.

I need this, for example, to set the title of the window that contains
a frame. This is done by calling SetWindowTextW, which needs a LPWSTR
in UCS-2/UTF-16, but the name to be set is received as a Lisp_Object
(Lisp_String?).

The function I'm changing is x_set_name, in w32fns.c.

>
>  . Converting a Lisp string into UTF-16 (or any other encoding) is an
>    _encode_ operation, so the last argument to
>    code_convert_string_norecord should be non-zero

You're right, thanks.
>

The issue was that I was using Qutf_16, which I believe adds a BOM
(and is big endian?). If I use Qutf_16le instead (no BOM,
little-endian) it's working ok.

Are there any issues that I need to worry about if I just call this
function everywhere this conversion is needed? e.g memory-leaks?