[Insight-developers] Re: [Insight-users] Fixing bugs for ITK with MinGW

[David Cole]

This is the fix that I checked in:
http://www.itk.org/cgi-bin/viewcvs.cgi/Utilities/itkjpeg/jconfig.h.in?root=Insight&sortby=date&r2=1.3&r1=1.2
(it's not the same as your patch, but it should work for you, too, I think...)

We have MinGW 5.13 stable release running on dash14. What version of
MinGW are you still having problems with...?


HTH,
David


On 8/31/07, Tomáš Kazmar <Tomash.Kazmar at seznam.cz> wrote:
> I do not have access to my mingw machine now, so I did not try to compile but
> it seems the patch was not applied to cvs. (or was it and i don't see it?)
>
> Which mingw version uses that machine from dashboard? As I wrote before,
> the bug is specific to recent versions of mingw.
>
> Tomas
>
> # But this has been fixed in CVS HEAD of ITK... This dashboard runs nightly at
> # Kitware:
> # http://www.itk.org/Testing/Sites/dash14.kitware/Win32-mingw/20070830-0100-Nightly/Test.html
> #
> # There are 21 failing tests, but no build errors or warnings with the MINGW
> # compiler.
> #
> # Are you sure you are using up to date source from CVS?
> #
> #
> #
> # On 8/30/07, Tomáš Kazmar <Tomash.Kazmar at seznam.cz> wrote:
> # >
> # > # Unfortunately, ITK has this little unresolved bug with MinGW. Could
> # > someone with
> # > # CVS access apply this simple patch:
> # > #
> # > # In /Utilities/itkjpeg/jmorecfg.h.in, somewhere before line 286, this
> # > should be
> # > # added:
> # > #
> # > # -----------------------------------------------
> # > # #if defined(__MINGW32__) && defined(_RPCNDR_H) && !defined(HAVE_BOOLEAN)
> # > # #define HAVE_BOOLEAN
> # > # #endif
> # > # -----------------------------------------------
> # >
> # > We discussed precisely this issue which arised in MinGW with win32api ver.
> # > >= 3.8.
> # > As it is a problem with win32api header files, it is present in both MinGW
> # > and Cygwin.
> # > I already sent a patch to itk-developers mailing list but it seems it was
> # > not applied to
> # > cvs. Could someone please check it in?
> # >
> # > Regards,
> # >
> # > Tomas
> # >
> # >
> # > I attach the patch, as well as the discussion from June 18th:
> # >
> # > #To avoid confusion: new version of win32api's rpcndr.h contains
> # > definitions of both
> # > #_RPCNDR_H, __RPCNDR_H__ and a typedef for boolean whereas old version
> # > #contains only __RPCNDR_H__.
> # > #
> # > #Tomas
> # > #
> # > ## ------------ Původní zpráva ------------
> # > ## Od: Mathieu Malaterre <mathieu.malaterre at gmail.com>
> # > ## Předmět: Re: patch to compile ITK2.8.0 on current MinGW/Cygwin
> # > (w32api>=3.8)
> # > ## Datum: 18.6.2007 14:18:16
> # > ## ----------------------------------------
> # > ## Forwarding to Brad King. Since he is the original author of the patch.
> # > ##
> # > ## Brad, looks like on MinGW __RPCNDR_H__ is not defined as expected after
> # > all.
> # > ##
> # > ## Thanks
> # > ## -Mathieu
> # > ##
> # > ## On 6/18/07, Tomáš Kazmar <Tomash.Kazmar at seznam.cz> wrote:
> # > ## > Hi,
> # > ## >
> # > ## > I do not know where to submit a patch for a bug that breaks
> # > compilation of
> # > ## > ITK2.8.0 on current MinGW/Cygwin, so I decided to send it to you
> # > directly.
> # > ## > If I am to submit the patch elsewhere, please tell me.
> # > ## >
> # > ## > In current MinGW/Cygwin there was added a definition of boolean and
> # > ## > a "symbol already defined" error is thrown for
> # > ## ITK/Utilities/itkjpeg/jmorecfg.h.
> # > ## > I suggest to patch jconfig.h.in like this:
> # > ## >
> # > ## > 24c24,25
> # > ## > < #if defined(_WIN32) && !(defined(__CYGWIN__) ||
> # > defined(__MINGW32__))
> # > ## > ---
> # > ## > > #if defined(_WIN32) && !((defined(__CYGWIN__) ||
> # > defined(__MINGW32__)) \
> # > ## > > && !defined(_RPCNDR_H))
> # > ## >
> # > ## > This is not a really nice solution but the only one that works.
> # > Because
> # > ## w32api.h
> # > ## > is not included in jconfig.h.in, we could not test for something like
> # > ## > W32API_VERSION < 308, neither there is any useful version information
> # > in
> # > ## > rpcndr.h itself (file where boolean is newly defined), the only thing
> # > left is
> # > ## to test for
> # > ## > !defined(_RPCNDR_H) which is coincidentally introduced in the same
> # > version of
> # > ## > rpcndr.h as the typedef.
> # > ## >
> # > ## >
> # > ## > Regards,
> # > ## > Tomas
> # > ## >
> # > ##
> # > ##
> # > ## --
> # > ## Mathieu
> # > _______________________________________________
> # > Insight-developers mailing list
> # > Insight-developers at itk.org
> # > http://www.itk.org/mailman/listinfo/insight-developers
> # >
> # >
> # >
> #
> #
> #
>