I did another test; I switched to mingw64 toolchain and did a 32-bit build,
same result: no build errors. The new gcc reports the following version:

c:\unicon>gcc --version
gcc (tdm64-2) 4.8.1

--Jafar


On Fri, Nov 6, 2015 at 12:35 AM, Jafar Al-Gharaibeh <[email protected]>
wrote:

> Don,
>
>    I fixed rusage related errors a while back, I discovered that I didn't
> commit this fixes - I just did.
>
> I checked out a fresh copy of Unicon sources and tried to do a 32-bit
> build. I have mingw32 for that with gcc 4.7.2. I then did
>
>   make W-Configure-GCC    # or NT-Configure-GCC
>
> You only need to do this one time only; the very first time after a fresh
> checkout to get windows makefile at the top level.
>
> I then did this:
>
>   make WUnicon32
>
> The build went smooth without any errors!   I then moved make.exe to
> mymake.exe and made sure I don't have any other make.exe on my path.
>
> I then did:
>
> mymake.exe WUnicon32
>
> To catch all hard-coded calls to make, to my surprise there was non on the
> build path. However there were a few MAKE variables that I have to change
> manually in these files:
>
> unicon/config/win32/gcc/makefile.top     # two occurrences at the top
> unicon/config/win32/gcc/makefile.wop   # two occurrences at the top
> unicon/config/win32/gcc/makedefs.top  # one occurrences at line 55
>
> in all cases I change
>
> MAKE=make
>
> to
>
> MAKE=mymake
>
> The build went smooth again with my "new" make with no errors.
>
> I haven't tried to enable jpg/png/threads yet, but at least until this
> point everything seems to be in order. Can you reproduce this? if not then
> it must be a difference in the the toolchain which is something we
> experienced many times in the past.
>
> Cheers,
> Jafar
>
>
>
>
> On Thu, Nov 5, 2015 at 4:47 PM, Don Ward <[email protected]> wrote:
>
>>
>> On 5 Nov 2015, at 21:57, Jafar Al-Gharaibeh <[email protected]> wrote:
>>
>>> The only way I’ve got it to work is if I have “make.exe’ available
>>> somewhere in my PATH. I also had to have the CPPFLAGS and CXX arguments. Do
>>> you build successfully without them?
>>> For me, without CXX I get g++ (which doesn’t go very well) and without
>>> the CPPFLAGS it can’t find ndbm.h (when compiling rttparse.c) and tp.h
>>> (when compiling libtp).
>>>
>>
>> Usually no, I don't need this, but the last time a tried to build on
>> Windows  a couple of weeks ago I had similar issues. There has been a
>> stream of commits to svn lately that shuffled few things around. We will
>> fix that
>>
>>
>> I was building rev 3976 (so I could compare what I had built with my
>> toolchain with the “official release” and hence validate my local build
>> environment).  I don’t see how recent svn commits could have affected rev
>> 3976, but I still need CPPFLAGS and CXX to build it.
>>
>> Part of the "make" problem -aside from the hardcoded calls to make-   is
>> that different "make" programs behave differently. Add differences between
>> different Windows releases/updates and you get a very unpredictable
>> behavior.  Last year, Clint took on a quest to build our own make (lives in
>> unicon/uni/umake.icn) to avoid some of this mess. Not sure if we are going
>> to switch to that at some point if it is mature enough.
>>
>>
>> If you do switch you’ll have the usual bootstrapping problem when
>> starting from scratch with a new build: you need uMake to be working in
>> order to build itself.
>>
>> I will rebuild from scratch on Windows in the next couple of days and fix
>> these new issues. I will let you know when I get that done.
>>
>>
>> OK (and thanks for all your efforts).  BTW, when I tried building rev
>> 4186, I got a stack of errors to do with rusage (which looks to have been
>> implemented circa rev 4052). Is this also a known problem for Windows
>> builds?
>>
>> Don
>>
>>
>
------------------------------------------------------------------------------
_______________________________________________
Unicon-group mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/unicon-group

Reply via email to