Home / Browse / MinGW - Minimalist GNU for Windows / Mail / Archive

MinGW - Minimalist GNU for Windows

Email Archive: mingw-msys (read-only)

[Mingw-msys] source code?

From: <selinger@ma...> - 2008-02-13 03:21

Hi,

I apologize in advance if I have overlooked the obvious answer to my
question: Where can I get the source code for the msysDVLPR package?

I have found the binary package at
http://sourceforge.net/project/showfiles.php?group_id=2435, and it is
essentially a tarball full of headers, executables, and libraries. But
unlike all the other MSYS/MinGW packages, there is no corresponding
*src* tarball.

Googling has not helped. I also browsed the CVS repository at
http://mingw.cvs.sourceforge.net/mingw/, but could not find the
pertinent sources.

In particular, I would like to see the sources for libmsys-1.0.dll.a.
More specifically, I would like to know how ioctl() and the termios
stuff has been implemented.

Thanks, -- Peter

Thread View

Thread	Author	Date
[Mingw-msys] source code?	<selinger@ma...>

Re: [Mingw-msys] source code?

From: Brian Dessent <brian@de...> - 2008-02-13 03:36

Peter Selinger wrote:

> Googling has not helped. I also browsed the CVS repository at
> http://mingw.cvs.sourceforge.net/mingw/, but could not find the
> pertinent sources.

It's there.  The main portion of the DLL is at
<http://mingw.cvs.sourceforge.net/mingw/msys/rt/src/winsup/cygwin/>. 
Remember that MSYS is a fork of an old version of Cygwin so the
directory layout and most of the files' descriptions will say that
they're part of Cygwin.

> In particular, I would like to see the sources for libmsys-1.0.dll.a.
> More specifically, I would like to know how ioctl() and the termios
> stuff has been implemented.

Err, that's an import library.  There's no code in that.  I think you
mean the sources for msys-1.0.dll.

Brian

Re: [Mingw-msys] floating point bug

From: <selinger@ma...> - 2008-04-14 22:11

Hi,

I am not sure if this list is the correct place for bug reports.  
I found the following code in the floating point library
newlib/libm/mathfp/sf_isinf.c:

int isinff (float x)
{
  __uint32_t wx;
  int exp;

  GET_FLOAT_WORD (wx, x);
  exp = (wx & 0x7f800000) >> 23;

  if ((exp == 0x7f8) && !(wx & 0xf0000))
    return (1);
  else
    return (0);
}

Note that exp can never equal 0x7f8, so this function always returns 0.
What was probably meant was either >> 20, or exp == 0xff.

-- Peter

Re: [Mingw-msys] floating point bug

From: Brian Dessent <brian@de...> - 2008-04-15 05:08

Peter Selinger wrote:

> I am not sure if this list is the correct place for bug reports.

This is code from newlib, and the version of newlib that MSYS includes
is extremely old, from 2001-09-07 as far as I can tell.  Maybe at some
point it would be a good idea to import a more fresh copy of newlib into
the tree.

> I found the following code in the floating point library
> newlib/libm/mathfp/sf_isinf.c:
> 
> int isinff (float x)
> {
>   __uint32_t wx;
>   int exp;
> 
>   GET_FLOAT_WORD (wx, x);
>   exp = (wx & 0x7f800000) >> 23;
> 
>   if ((exp == 0x7f8) && !(wx & 0xf0000))
>     return (1);
>   else
>     return (0);
> }

Current newlib has this:

int
_DEFUN (isinff, (x),
        float x)
{
        __int32_t ix;
        GET_FLOAT_WORD(ix,x);
        ix &= 0x7fffffff;
        return FLT_UWORD_IS_INFINITE(ix);
}

It looks like this change was made when the file was moved from mathfp/
to common/ around 2006-02-26:

<http://sourceware.org/cgi-bin/cvsweb.cgi/src/newlib/libm/common/sf_isinff.c?rev=1.1&content-type=text/x-cvsweb-markup&cvsroot=src>
<http://sourceware.org/ml/newlib/2006/msg00132.html>

Brian