[CMake] Bug #12189

[aaron.meadows at thomsonreuters.com]

I've updated the bug with an additional patch (had a  bug for some
configuration cases).

 

What is the procedure to get this patch considered and applied?  Do I
need to contact the Visual Studio Generator maintainer directly?

 

Aaron C. Meadows 

From: Meadows, Aaron C. 
Sent: Tuesday, June 21, 2011 9:49 AM
To: Meadows, Aaron C.; cmake at cmake.org
Subject: RE: [CMake] Bug #12189

 

I updated the bug with this patch and some of the details from below.

 

Aaron C. Meadows 

From: cmake-bounces at cmake.org [mailto:cmake-bounces at cmake.org] On Behalf
Of Meadows, Aaron C.
Sent: Monday, June 20, 2011 5:17 PM
To: cmake at cmake.org
Subject: [CMake] Bug #12189

 

(link: http://public.kitware.com/Bug/view.php?id=12189)

 

I just came across the above bug opened last month,  Summaraized as: "It
is not possible to generate a Visual Studio project with ASCII/SBCS
character set".  (Full Bug Text Following Message)

 

I find myself in the situation where I need this to be set to "not set",
so as to have ASCII/SBCS as the character set for all my projects.  I
dug back through the last few years of the maillist and didn't see
anything about this issue.   

 

Anyone have experience with this issue?  And better yet, know of a
solution?

 

I checked the source file he indicated (it's on line 702 of the 2.8.4
source):

  // If unicode is enabled change the character set to unicode, if not

  // then default to MBCS.

  if(targetOptions.UsingUnicode())

    {

    fout << "\t\t\tCharacterSet=\"1\">\n";

    }

  else

    {

    fout << "\t\t\tCharacterSet=\"2\">\n";

    }

 

Additionally, VS 2010 has a different mechanism (line 287 of the 2.8.4
source):

    if(this->Target->GetType() <= cmTarget::MODULE_LIBRARY &&

       this->ClOptions[*i]->UsingUnicode())

      {

      this->WriteString("<CharacterSet>Unicode</CharacterSet>\n", 2);

      }

    else

      {

      this->WriteString("<CharacterSet>MultiByte</CharacterSet>\n", 2);

      }

 

In the case of VS 2010, the proper ASCII setting is:

       <CharacterSet>NotSet</CharacterSet>

 

 

It would be fairly trivial to change both these cases to generate the
"Not Set" case.  However, how should backward compatibility be
maintained, if at all?  

 

I've created and attached a patch which resolves this bug by adding an
"_SBCS" define which is checked.  If it is set and "_UNICODE" is not
set, it will write the correct files for the "Not Set" ASCII case.  This
is slightly different than the suggestion in the Bug, but was very easy
to write and provides identical behavior if "_SBCS" is not set.

 

 

------------------------------------------------------------------------
----------------------------------------------------------

Full Bug Text:

------------------------------------------------------------------------
----------------------------------------------------------

In Visual Studio 9.0 (and prior, 10.0 i don't know) it is possible
specify three different character sets for your project within the
project properties:

Not Set = ASCII/SBCS (Single Byte Character Set)
Unicode
Multi-Byte

Depending on the option different preprocessor defines are set
(http://msdn.microsoft.com/en-us/library/c426s321(v=vs.80).aspx [^
<http://msdn.microsoft.com/en-us/library/c426s321(v=vs.80).aspx> ]):

SBCS: neither _UNICODE nor _MBCS defined
Unicode: _UNICODE defined
Multi_Byte: _MBCS defined

The character set settings is stored within the vs proj files as an xml
attribute:

SBCS: CharacterSet="0"
Unicode: CharacterSet="1"
Multi-Byte: CharacterSet="2"

However, the cmake visual studio generators do not support generating of
projects with CharacterSet="0" (see cmLocalVisualStudio7Generator.cxx
line 730). At the moment the generators select unicode if a _UNICODE
macro has been set by add_definitions, otherwise multi-byte is selected.

To solve the problem and to keep backwards compatability, i suggest to
define the _MBCS macro by default for the visual studio generators and
to set CharacterSet="2" only if this macro is still available and
otherwise CharacterSet="0". In that case the user can remove the _MBCS
macro by remove_definitions and select this way the SBCS. If the user
adds _UNICODE by add_definitions CharacterSet="1" should be selected and
the conflicting _MBCS macro must be deleted by the code generator.

------------------------------------------------------------------------
----------------------------------------------------------

 

Aaron Meadows
Software Engineer

Thomson Reuters

Phone: 314.468.3530
Mobile: 636.541.6139
aaron.meadows at thomsonreuters.com
thomsonreuters.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.cmake.org/pipermail/cmake/attachments/20110623/e6220e1a/attachment.htm>