Paul Burba wrote:
> On Mon, Nov 23, 2009 at 10:00 AM, Branko Čibej <br...@xbc.nu> wrote:
>   
>> Paul Burba wrote:
>>     
>>> On Sun, Nov 22, 2009 at 9:49 PM, Arfrever Frehtes Taifersar Arahesis
>>> <arfrever....@gmail.com> wrote:
>>>
>>>       
>>>> 2009-11-19 16:41:43 Paul Burba napisał(a):
>>>>
>>>>         
>>>>> On Mon, Nov 16, 2009 at 2:16 PM, Paul Burba <ptbu...@gmail.com> wrote:
>>>>>
>>>>>           
>>>>>> On Sat, Nov 14, 2009 at 2:42 PM, Arfrever Frehtes Taifersar Arahesis
>>>>>> <arfrever....@gmail.com> wrote:
>>>>>>
>>>>>>             
>>>>>>> Author: arfrever
>>>>>>> Date: Sat Nov 14 11:42:26 2009
>>>>>>> New Revision: 40510
>>>>>>>
>>>>>>> Log:
>>>>>>> Move inclusion of svn_debug.h from svn_types.h to svn_private_config.h.
>>>>>>>
>>>>>>> * subversion/include/svn_types.h: Don't include "private/svn_debug.h".
>>>>>>>
>>>>>>> * configure.ac: Include "private/svn_debug.h" in generated 
>>>>>>> svn_private_config.h.in
>>>>>>>   file.
>>>>>>>
>>>>>>> * subversion/svn_private_config.hw: Include "private/svn_debug.h".
>>>>>>>
>>>>>>> Review by: gstein
>>>>>>>
>>>>>>> Modified:
>>>>>>>   trunk/configure.ac
>>>>>>>   trunk/subversion/include/svn_types.h
>>>>>>>   trunk/subversion/svn_private_config.hw
>>>>>>>
>>>>>>> Modified: trunk/configure.ac
>>>>>>> URL: 
>>>>>>> http://svn.collab.net/viewvc/svn/trunk/configure.ac?pathrev=40510&r1=40509&r2=40510
>>>>>>> ==============================================================================
>>>>>>> --- trunk/configure.ac  Sat Nov 14 00:02:24 2009        (r40509)
>>>>>>> +++ trunk/configure.ac  Sat Nov 14 11:42:26 2009        (r40510)
>>>>>>> @@ -615,7 +615,15 @@ if test "$enable_nls" = "yes"; then
>>>>>>>   fi
>>>>>>>  fi
>>>>>>>
>>>>>>> -AH_BOTTOM(
>>>>>>> +AH_BOTTOM([/*
>>>>>>> + * Subversion developers may want to use some additional debugging 
>>>>>>> facilities
>>>>>>> + * while working on the code. We'll pull that in here, so individual 
>>>>>>> source
>>>>>>> + * files don't have to include this header manually.
>>>>>>> + */
>>>>>>> +#ifdef SVN_DEBUG
>>>>>>> +#include "private/svn_debug.h"
>>>>>>> +#endif
>>>>>>> +
>>>>>>>  #define N_(x) x
>>>>>>>  #define U_(x) x
>>>>>>>  #ifdef ENABLE_NLS
>>>>>>> @@ -629,7 +637,7 @@ AH_BOTTOM(
>>>>>>>  #define gettext(x) (x)
>>>>>>>  #define dgettext(domain, x) (x)
>>>>>>>  #endif
>>>>>>> -)
>>>>>>> +])
>>>>>>>
>>>>>>>  dnl Used to simulate makefile conditionals.
>>>>>>>  GETTEXT_CODESET=\#
>>>>>>>
>>>>>>> Modified: trunk/subversion/include/svn_types.h
>>>>>>> URL: 
>>>>>>> http://svn.collab.net/viewvc/svn/trunk/subversion/include/svn_types.h?pathrev=40510&r1=40509&r2=40510
>>>>>>> ==============================================================================
>>>>>>> --- trunk/subversion/include/svn_types.h        Sat Nov 14 00:02:24 
>>>>>>> 2009        (r40509)
>>>>>>> +++ trunk/subversion/include/svn_types.h        Sat Nov 14 11:42:26 
>>>>>>> 2009        (r40510)
>>>>>>> @@ -1144,14 +1144,4 @@ typedef unsigned long svn_linenum_t;
>>>>>>>  #include "svn_error.h"
>>>>>>>
>>>>>>>
>>>>>>> -/*
>>>>>>> - * Subversion developers may want to use some additional debugging 
>>>>>>> facilities
>>>>>>> - * while working on the code. We'll pull that in here, so individual 
>>>>>>> source
>>>>>>> - * files don't have to include this header manually.
>>>>>>> - */
>>>>>>> -#ifdef SVN_DEBUG
>>>>>>> -#include "private/svn_debug.h"
>>>>>>> -#endif
>>>>>>> -
>>>>>>> -
>>>>>>>  #endif /* SVN_TYPES_H */
>>>>>>>
>>>>>>> Modified: trunk/subversion/svn_private_config.hw
>>>>>>> URL: 
>>>>>>> http://svn.collab.net/viewvc/svn/trunk/subversion/svn_private_config.hw?pathrev=40510&r1=40509&r2=40510
>>>>>>> ==============================================================================
>>>>>>> --- trunk/subversion/svn_private_config.hw      Sat Nov 14 00:02:24 
>>>>>>> 2009        (r40509)
>>>>>>> +++ trunk/subversion/svn_private_config.hw      Sat Nov 14 11:42:26 
>>>>>>> 2009        (r40510)
>>>>>>> @@ -91,3 +91,12 @@
>>>>>>>  #define APU_WANT_DB
>>>>>>>  #include <apu_want.h>
>>>>>>>  #endif
>>>>>>> +
>>>>>>> +/*
>>>>>>> + * Subversion developers may want to use some additional debugging 
>>>>>>> facilities
>>>>>>> + * while working on the code. We'll pull that in here, so individual 
>>>>>>> source
>>>>>>> + * files don't have to include this header manually.
>>>>>>> + */
>>>>>>> +#ifdef SVN_DEBUG
>>>>>>> +#include "private/svn_debug.h"
>>>>>>> +#endif
>>>>>>>
>>>>>>>               
>>>>>> Hi Arfrever,
>>>>>>
>>>>>> This change breaks the debug build on Windows since __attribute__ is
>>>>>> not longer defined, see __attribute__no.longer.defined.txt.
>>>>>>
>>>>>> I thought that simply adding #include <apr.h> to svn_debug.h would fix
>>>>>> this, but I end up with even more weirdness, see include.apr.h.txt.
>>>>>>
>>>>>> I hate to admit I can't quite figure this out.  I've reverted this
>>>>>> change locally so I can build, but please look into this when you can.
>>>>>>
>>>>>>             
>>>>> Hi Arfrever,
>>>>>
>>>>> I've been unable to fix the broken Windows debug build and it doesn't
>>>>> appear anyone else is working on this.  Edmund has confirmed the
>>>>> breakage on a his machine.  So, per our rules in HACKING, I
>>>>> regrettably reverted this change in r882182.
>>>>>
>>>>> I'm more than happy to test a new version of r40510(r880584) or assist
>>>>> in any way I can.
>>>>>
>>>>>           
>>>> Please test the attached patch.
>>>>
>>>>         
>>> Hi Arfrever,
>>>
>>> That works.  I still don't understand why we can't use __attribute__
>>> in svn_debug.h.  It is used in svn_cmdline.h, svn_io.h, svn_string.h,
>>> and various other places without any problems.
>>>
>>>       
>> Perhaps svn_debug.h doesn't include apr.h, where __attribute__ is
>> defined for compilers other than GCC?
>>     
>
> Hi Brane,
>
> That was my initial thought too and I tried it, but received some very
> odd results.  See the start of this thread:
> http://mail-archives.apache.org/mod_mbox/subversion-dev/200911.mbox/browser.
>   

Hmm, curious ... I'd suggest producing a preprocessed source to see
where apr.h is first included. From looking at the surces, I can't see
what would be going wrong, either.

-- Brane

Reply via email to