Use of the L prefix for wide strings in resource files

[Paul Vriens]

Hi,

Yaron just made me aware of something he and I discussed briefly a few 
weeks ago.

The shlwapi resource files uses the following:

PUSHBUTTON L"&OK", IDOK, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
  PUSHBUTTON L"&Cancel", IDCANCEL, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
  PUSHBUTTON L"&Yes", IDYES, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
  PUSHBUTTON L"&No", IDNO, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP

There doesn't seem to be an issue with most of the languages as our 
resource compiler seem to do things just right.

I did however see strangeness for some languages and also the not yet 
submitted Hebrew translation (maybe more?):

http://source.winehq.org/transl/resource.php?branch=master&lang=011%3A01...
http://source.winehq.org/transl/resource.php?branch=master&lang=022%3A01...

ResHacker shows the same strangeness so it appears not only to be the 
translation site.

Any idea where to start looking?

[Reece Dunn]

On 5 December 2010 14:09, Paul Vriens  wrote:
> Hi,
>
> Yaron just made me aware of something he and I discussed briefly a few weeks
> ago.
>
> The shlwapi resource files uses the following:
>
> PUSHBUTTON L"&OK", IDOK, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
>  PUSHBUTTON L"&Cancel", IDCANCEL, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
>  PUSHBUTTON L"&Yes", IDYES, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
>  PUSHBUTTON L"&No", IDNO, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
>
> There doesn't seem to be an issue with most of the languages as our resource
> compiler seem to do things just right.
>
> I did however see strangeness for some languages and also the not yet
> submitted Hebrew translation (maybe more?):
>
> http://source.winehq.org/transl/resource.php?branch=master&lang=011%3A01...
> http://source.winehq.org/transl/resource.php?branch=master&lang=022%3A01...
>
> ResHacker shows the same strangeness so it appears not only to be the
> translation site.
>
> Any idea where to start looking?

The Cancel button in Japanese looks like raw ShiftJIS in windows 1252
or utf8 encoding.

What appears to be happening is:
  1/  the localised strings are being converted to their multi-byte
representation (e.g. ShiftJIS for the Japanese (lang=011) text)
  2/  the multi-byte string is being encoded in the resource codepage (utf8?)

This would make sense if the L"..." string is going through
MultiByteToWideCharacter with the ShiftJIS encoding (for Japanese) and
then the resource is assuming this is utf8 (or whatever the resource
file codepage is) and is going through MultiByteToWideCharacter for
that as well.

I can't recognise the latin character sequences in the Cyrillic
dialogs (lang 022); these may be the utf8 sequences.

- Reece

[Alexandre Julliard]

Paul Vriens  writes:
> Hi,
>
> Yaron just made me aware of something he and I discussed briefly a few
> weeks ago.
>
> The shlwapi resource files uses the following:
>
> PUSHBUTTON L"&OK", IDOK, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
>  PUSHBUTTON L"&Cancel", IDCANCEL, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
>  PUSHBUTTON L"&Yes", IDYES, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
>  PUSHBUTTON L"&No", IDNO, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
>
> There doesn't seem to be an issue with most of the languages as our
> resource compiler seem to do things just right.
>
> I did however see strangeness for some languages and also the not yet
> submitted Hebrew translation (maybe more?):
>
> http://source.winehq.org/transl/resource.php?branch=master&lang=011%3A01...
> http://source.winehq.org/transl/resource.php?branch=master&lang=022%3A01...
>
> ResHacker shows the same strangeness so it appears not only to be the
> translation site.
>
> Any idea where to start looking?

L"" strings are simply expanded from 8-bit to 16-bit, so it's only
suitable for strings encoded in Latin-1. In fact the parser should
probably reject anything that doesn't fit in 7-bit ASCII.

[Paul Vriens]

On 12/05/2010 04:14 PM, Alexandre Julliard wrote:
> Paul Vriens  writes:
> > Hi,
> >
> > Yaron just made me aware of something he and I discussed briefly a few
> > weeks ago.
> >
> > The shlwapi resource files uses the following:
> >
> > PUSHBUTTON L"&OK", IDOK, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
> >   PUSHBUTTON L"&Cancel", IDCANCEL, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
> >   PUSHBUTTON L"&Yes", IDYES, 105, 40, 50, 14, WS_GROUP | WS_TABSTOP
> >   PUSHBUTTON L"&No", IDNO, 160, 40, 50, 14, WS_GROUP | WS_TABSTOP
> >
> > There doesn't seem to be an issue with most of the languages as our
> > resource compiler seem to do things just right.
> >
> > I did however see strangeness for some languages and also the not yet
> > submitted Hebrew translation (maybe more?):
> >
> > http://source.winehq.org/transl/resource.php?branch=master&lang=011%3A01...
> > http://source.winehq.org/transl/resource.php?branch=master&lang=022%3A01...
> >
> > ResHacker shows the same strangeness so it appears not only to be the
> > translation site.
> >
> > Any idea where to start looking?
>
> L"" strings are simply expanded from 8-bit to 16-bit, so it's only
> suitable for strings encoded in Latin-1. In fact the parser should
> probably reject anything that doesn't fit in 7-bit ASCII.

So we can have a patch that removes those L"" strings and one to change 
the behavior of the parser?

I guess we can get safely rid of all L"" strings instead of just the 
ones that show the issue.