Re: [Mingw-users] Program Distribution - Installers (win32/unix)

[Peter Mcleod <peter_m@an...>]

Thanks heaps - I guess I'll use the pascal setup utility
for a self extracting win32 install and then just learn how
to create a few of the popular linux packages.  I'll look
into mac-osx a little more carefully later then - when I've
at least got a test system (need to upgrade from mac-os9
first).  

For  those of you interested I'm currently testing my own
program installations on:

windows 98, windows NT, red hat linux, yellow dog linux (I
also have mac-os9).  As I'm not doing anything out of the
ordinary I suspect that's enough to expect other versions
of windows and linux to install correctly.

On Wed, 21 May 2003 06:48:25 -0400
 Earnie Boyd <earnie_boyd@...> wrote:
> Peter Mcleod wrote:
> >Hi all,
> >
> >When it comes to installation programs, a preference I
> have
> >always had is for a friendly self-extracting
> (compressed)
> >executable that automatically copies its contents to a
> temp
> >directory and then starts up an install program (this is
> >often helpful as there are still many computer users out
> >there who are not so literate).  The MinGW and MSYS
> >binaries themselves are provided in such a download
> format.
> > 
> 
> Thanks, I agree.
> 
> >
> >My questions are:
> >
> >1.  Ideally, I would like to package my cross-os
> >programs/utilities like this - is there anything in the
> >minGW/gcc world for setting up something similar so that
> it
> >produces a win32 and a unix installer?  How did they
> >achieve it with the MinGW installer?  What was used for
> the
> >GUI - surely not GTK without a runtime?  --It may help
> to
> >know I always keep my source code 100% compatible
> between
> >win32 and unix - that is, no macros, no using os
> specific
> > function/run-time libraries.  I don't even write to the
> >windows registery but expect that the installer would
> have
> >to - for the windows "add/remove programs" utility.
> >
> 
> I use an open source utility named INNO Setup which you
> can find at http://www.jrsoftware.org/isinfo.php and a
> bit of bash scripting.  The drawback is that the source
> is Borland Pascal.  INNO Setup uses a script to control
> what is included and has pre and post install controls.
>  It doesn't have to write to registry at all but will if
> you choose to include uninstall procedures only for the
> "add/remove programs" utility.
> 
> >2.  My previous experience with install programs are all
> >win32 related - I understand the various linux
> >distributions have their own package managers (I've used
> >RPM quite a bit but that's all).  How does the linux
> world
> >co-ordinate installations (if at all)?  I wouldn't want
> to
> >lock out anybody from using any of my programs/utilities
> >because they don't install anything that doesn't go
> through
> >their package manager but at the same time I don't want
> to
> >create and maintain heaps and heaps of install packages?
>  
> >
> 
> What a delima, I would hope that package distributors
> using RPM and DEBIAN dpkg have some nice utilities to
> help that delima.  I don't know if you can get the INNO
> Setup to build for linux, I doubt it, it's mostly Win32
> oriented.  I do know that there are ports of some style
> of both RPM and DPKG for win32, there is even a
> debian-win32 list at debian.org.
> 
> >Whatever info you have, recommendations or sites that
> you
> >can point me to would be appreciated?  If I have to pay
> for
> >a utility (that supports win32 and linux) and it saves
> me
> >the bother of too much maintenance, I would be glad to.
>  
> >
> 
> Well, let me know what you come up with.  I certainly
> don't have an answer for something Linux.  INNO Setup
> works well for Win32.
> 
> >I have yet another question (which I hope won't detract
> >anybody from the above and more important to me) - and
> that
> >is, how much like linux/unix is mac-osx - do we have a
> gcc
> >compiler available for it - the standard compiler?  If I
> >write a plain ansi c program in linux and compile it
> will
> >it run on linux - or do I need to recompile it on that
> os?
> > 
> 
> IIRC, gcc does exist for OSX.  You will need to recompile
> and link your binaries unless emulation libraries are
> provided, such as WINE or LINE.
> 
> HTH,
> Earnie.
> 

 
==
Download ringtones, logos and picture messages at Ananzi Mobile Fun.
http://www.ananzi.co.za/cgi-bin/goto.pl?mobile

Re: [Mingw-users] Strange file size

[Oscar Fuentes <ofv@wa...>]

Anders Lind=E9n <mingw@...> writes:

> Hello!
>
> I have a strange situation.
>
> I have this files in my project:
>
> 2003-05-21  20:51               31 365 test.cpp
> 2003-05-21  23:13               49 533 test.o
> 2003-05-21  23:13              499 876 test.dll
> 2003-05-21  23:13                4 452 libstdll.a
> 2003-05-12  15:05                   87 test.def
> 2003-05-21  19:46                  218 Makefile
> 2003-05-16  23:02                   44 test.bat
> 2003-05-21  20:07                  311 file.txt
>
>
> The problem is the really big dll file.  I cant say what made the
> dll-file this big. It was something I did in the code that made it
> go from <100000 to 500000.  When I strip this dll I get a file that
> is 245760 bytes big, but thats nothing compared to the small dll I
> had previously. (It was below 50k).
>
>
> This is my makefile, maybe there is a clue in here:
>
> all:
>     g++ -c -DBUILD_DLL -fexceptions test.cpp -o test.o
>     dllwrap --output-lib=3Dlibstdll.a --dllname=3Dtest.dll --driver-name=
=3Dg++ --def=3Dtest.def --entry _DllMain --add_stdcall_alias
> test.o -lole32  -loleaut32 -lwsock32
>
> Am I doing this in the right way?

dllwrap is deprecated. gcc/g++ now supports the -shared switch, that
creates a dll instead of an exe. I don't think this is the culprit of
the size change.

> I cannot put the test.cpp file here, but since the compilation of it
> resulted in 49533 bytes, it feels very unlikely that the contents in
> the cpp-file can affect the dll size this much.
>
> The def-file however:
>
> EXPORTS
>         DllRegisterServer
>         DllUnregisterServer
>         DllGetClassObject
>         DllCanUnloadNow
>
>
>
> Any suggestions?
> What was magically causing this big dll size?

Just making use of some I/O Standard C++ Library functionality is
enough. A "hello world" c++ app using iostreams creates an executable
larger than >400KB with my g++ 3.2.3 installation.

> Any comments appreciated!
> Maybe reinstalling mingw can be it?

I don't think so.

--=20
Oscar

Re: [Mingw-users] compiler byte swapping

[Benjamin Riefenstahl <Benjamin.Riefenstahl@ep...>]

Hi Alan, Oscar,


Oscar Fuentes <ofv@...> writes:
> I think this is the problem:
>
> WRITELONG( '8BPS' );
>
> In C++, how the compiler handles multibyte character constants is
> implementation-defined. I guess the same applies to C.

Ah, it's about those.  Yes, what '8BPS' actually means here is more or
less undefined, neither its integer value nor its actual byte-order is
specified.  We know that it has type "int" or "long", but that's about
all we know from the language POV.

Unless of course the compiler makes it an extension to explicitly
document how it's supposed to work, like the compilers on MacOS do it.
But I don't believe gcc has ever made any garantees about this, at
least not on platforms other than MacOSX.

Alan, you should use char[4] instead, if you can do that.  Otherwise
you need to write special binary parsers/formatters just for these
kind of type codes, with their actual implementation depending on the
compiler.  Also make sure that they are separate from the
parsers/formatters for int/long.


so long, benny

[Mingw-users] Strange file size

[<mingw@ca...>]

Hello!

I have a strange situation.

I have this files in my project:

2003-05-21  20:51               31 365 test.cpp
2003-05-21  23:13               49 533 test.o
2003-05-21  23:13              499 876 test.dll
2003-05-21  23:13                4 452 libstdll.a
2003-05-12  15:05                   87 test.def
2003-05-21  19:46                  218 Makefile
2003-05-16  23:02                   44 test.bat
2003-05-21  20:07                  311 file.txt


The problem is the really big dll file.
I cant say what made the dll-file this big. It was something I did in the code that made it go from  <100000 to 500000.
When I strip this dll I get a file that is 245760 bytes big, but thats nothing compared to the small dll I had previously. (It was
below 50k).


This is my makefile, maybe there is a clue in here:

all:
    g++ -c -DBUILD_DLL -fexceptions test.cpp -o test.o
    dllwrap --output-lib=libstdll.a --dllname=test.dll --driver-name=g++ --def=test.def --entry _DllMain --add_stdcall_alias
test.o -lole32  -loleaut32 -lwsock32

Am I doing this in the right way?

I cannot put the test.cpp file here, but since the compilation of it resulted in 49533 bytes, it feels very unlikely that the
contents in the cpp-file can affect the
dll size this much.

The def-file however:

EXPORTS
        DllRegisterServer
        DllUnregisterServer
        DllGetClassObject
        DllCanUnloadNow



Any suggestions?
What was magically causing this big dll size?

Any comments appreciated!
Maybe reinstalling mingw can be it?


/Anders Lindén

[Mingw-users] Re: compiler byte swapping

[Oscar Fuentes <ofv@wa...>]

"Alan Smallbone" <asmallbone@...> writes:

> Hi Oscar,
>
> Thanks for responding, I have attached a zip file which I hacked
> bits of the program out of and made it into a short executable that
> illustrates the problem that I am seeing. It is not in the "mywrite"
> definition as I first thought, it appears to be in how the data is
> passed to it from the "WRITELONG" as defined in this program. In the
> zip file is the header file and the test case program, and please,
> it is not the best thing in the world as an example. Also included
> are results from the two versions of the c compiler the
> fwrite.out_test1_V1 is from version 2.95.3-6 of gcc and is in the
> correct order, the fwrite.out_test1_V2 file is from gcc version 3.2
> from the latest version of mingw. Thanks so much for the help.

[Please keep discussion on list, thanks.]

I think this is the problem:

WRITELONG( '8BPS' );

In C++, how the compiler handles multibyte character constants is
implementation-defined. I guess the same applies to C. Maybe gcc 2.9x
keeps the byte order and 3.x swaps it. Have you tried with integer
constants instead of multibyte character constants?

-- 
Oscar

Re: [Mingw-users] Program Distribution - Installers (win32/unix)

[Nicolay Giraldo <ngiraldo@co...>]

On Wed, 21 May 2003 11:29:12 -0400, Earnie Boyd <earnie_boyd@...> 
wrote:

>> Creating a installer is just creating a script and calling the nsis 
>> compiler, you can put that in a makefile if you want. Compile times for 
>> the installer are a matter of a couple of seconds.
>>
>
> Wouldn't that depend on the size of the distribution included in the 
> installer?
>

Mmm.  Yes.  It depends also in the compressor. When using bzip it's slower.


> Not necessary, but highly recommended.  However, if I found an installer 
> that worked well on both Win32 and a Linux flavor, I wouldn't be opposed 
> to using a different installer than what is considered standard for that 
> system.

Source code is in C++ (but full of HWND and other MSW stuff), it would be 
great to have a single linux installer. Even if it just detects the distro 
and then launches the package manager with the right package.


-- 
Nicolay Giraldo
------------------------
We are Turing O-Machines
The Oracle is out there

[Mingw-users] ssize_t on mingw

[Sascha Sommer <saschasommer@fr...>]

Is there a reason why there is no ssize_t type for mingw?
If not, could it please be added or do you prefere a patch?

Sascha

[Mingw-users] make 3.80

[Jed Roberts <jedroberts@po...>]

Hi,

I just downloaded the new make v3.80 and it doesn't read my makefile.  =
This works:

make -fmakefile all

But this doesn't:

make all

I get:
make: *** No rule to make target `all'.  Stop.

I had been using the 3.79.1 make from DJGPP and it did not have this =
problem.  What am I doing wrong?

Jed Roberts

Re: [Mingw-users] -mwindows, with printf output to Console that I start the program??

[Timothy Schepens <tischepe@fa...>]

Greg Chicares wrote:
> psksvp wrote:
> 
>>If I build a Win32 app with flag -mwindows, I wonder if there is a way to
>>enable console output from printf, to the console(sh) that I start the
>>program(started from commandline in msys sh)?
> 
> 
> IIRC, a pure GUI app can't have a console,
> but a console app can have a GUI.
> 
> 


A GUI app can get a console with AllocConsole, but it will create a new 
console window. So you will have to write a console app and add a GUI to it.

Timothy

Re: [Mingw-users] -mwindows, with printf output to Console that I start the program??

[TROCHU Xavier <xtrochu@ya...>]

--- Greg Chicares <chicares@...> wrote:
> psksvp wrote:
> > 
> > If I build a Win32 app with flag -mwindows, I
> wonder if there is a way to
> > enable console output from printf, to the
> console(sh) that I start the
> > program(started from commandline in msys sh)?
> 
> IIRC, a pure GUI app can't have a console,
> but a console app can have a GUI.
> 

Not completly correct.

You can create a console with a GUI application, using
AllocConsole() but :
_ It is a new console, not the one where you start the
program.
_ You can't directly call printf in it. You have to
link the windows HANDLE to mscrt handle then to stdio
FILE *.

Looking at MSDN, you can call AttachConsole() to
attach your process to the console of another process,
but you have to know it's Process ID. And it's only
available for XP.

Xavier


__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com

Re: [Mingw-users] -mwindows, with printf output to Console that I start the program??

[Earnie Boyd <earnie_boyd@ya...>]

psksvp wrote:
> Hello
> 
> If I build a Win32 app with flag -mwindows, I wonder if there is a way to
> enable console output from printf, to the console(sh) that I start the
> program(started from commandline in msys sh)?
> 

No, but you could use CreateConsole and do that magic to open an output 
window.

Earnie.

Re: [Mingw-users] -mwindows, with printf output to Console that I start the program??

[Greg Chicares <chicares@mi...>]

psksvp wrote:
> 
> If I build a Win32 app with flag -mwindows, I wonder if there is a way to
> enable console output from printf, to the console(sh) that I start the
> program(started from commandline in msys sh)?

IIRC, a pure GUI app can't have a console,
but a console app can have a GUI.

Re: [Mingw-users] Program Distribution - Installers (win32/unix)

[Earnie Boyd <earnie_boyd@ya...>]

Nicolay Giraldo wrote:
> 
> I think the best windows installer I've seen and used is NSIS, it's made 
> by nullsoft, the ones who made winamp. NSIS is opensource, well it's 
> more public domain than BSD/GPL.
> 

Looks interesting.

> Creating a installer is just creating a script and calling the nsis 
> compiler, you can put that in a makefile if you want. Compile times for 
> the installer are a matter of a couple of seconds.
> 

Wouldn't that depend on the size of the distribution included in the 
installer?

> I was looking at the Inno Setup page and the faq, and a lot of the "not 
> possible now" features of Inno Setup, like changing the icon, 
> conditional instalation and /silent installs are easy with NSIS installer.
> 

I plan to look at NSIS for comparison, but I could create makefile 
targets for INNO Setup as well.

> It doesn't create unix installers, I think that for linux at least is 
> necessary to create a package for the specific linux distro, i.e. 
> Mandrake 8.x RPM, Red Hat 7.x RPM, Debian .deb, and so on.
> 

Not necessary, but highly recommended.  However, if I found an installer 
that worked well on both Win32 and a Linux flavor, I wouldn't be opposed 
to using a different installer than what is considered standard for that 
system.

Earnie.

Re: [Mingw-users] -mwindows, with printf output to Console that I start the program??

[Benjamin Riefenstahl <Benjamin.Riefenstahl@ep...>]

Hi,


"psksvp" <psksvp@...> writes:
> If I build a Win32 app with flag -mwindows, I wonder if there is a
> way to enable console output from printf, to the console(sh) that I
> start the program(started from commandline in msys sh)?

No.

You can compile without -mwindows and still create a GUI, if that fits
the bill.  You will need to add -lgdi32 if you explicitly use drawing
functions.  You can also use the "run" utility from the Xfree86
package to hide the console, for cases where you don't want it, like
for desktop shortcuts.


Hope this helps, benny

[Mingw-users] -mwindows, with printf output to Console that I start the program??

[psksvp <psksvp@cc...>]

Hello

If I build a Win32 app with flag -mwindows, I wonder if there is a way to
enable console output from printf, to the console(sh) that I start the
program(started from commandline in msys sh)?

psksvp

Re: [Mingw-users] Program Distribution - Installers (win32/unix)

[Nicolay Giraldo <ngiraldo@co...>]

On Wed, 21 May 2003 08:55:37 +0200, Peter Mcleod <peter_m@...> 
wrote:

> Hi all,
>
> When it comes to installation programs, a preference I have
> always had is for a friendly self-extracting (compressed)
> executable that automatically copies its contents to a temp
> directory and then starts up an install program (this is
> often helpful as there are still many computer users out
> there who are not so literate).  The MinGW and MSYS
> binaries themselves are provided in such a download format.
>
>
> My questions are:
>
> 1.  Ideally, I would like to package my cross-os
> programs/utilities like this - is there anything in the
> minGW/gcc world for setting up something similar so that it
> produces a win32 and a unix installer?  How did they
> achieve it with the MinGW installer?  What was used for the
> GUI - surely not GTK without a runtime?  --It may help to
> know I always keep my source code 100% compatible between
> win32 and unix - that is, no macros, no using os specific
> function/run-time libraries.  I don't even write to the
> windows registery but expect that the installer would have
> to - for the windows "add/remove programs" utility.

I think the best windows installer I've seen and used is NSIS, it's made by 
nullsoft, the ones who made winamp. NSIS is opensource, well it's more 
public domain than BSD/GPL.

Creating a installer is just creating a script and calling the nsis 
compiler, you can put that in a makefile if you want. Compile times for the 
installer are a matter of a couple of seconds.

I was looking at the Inno Setup page and the faq, and a lot of the "not 
possible now" features of Inno Setup, like changing the icon, conditional 
instalation and /silent installs are easy with NSIS installer.

It doesn't create unix installers, I think that for linux at least is 
necessary to create a package for the specific linux distro, i.e. Mandrake 
8.x RPM, Red Hat 7.x RPM, Debian .deb, and so on.


-- 
Nicolay Giraldo
------------------------
We are Turing O-Machines
The Oracle is out there

Re: [Mingw-users] Compiler changes

[Benjamin Riefenstahl <Benjamin.Riefenstahl@ep...>]

Hi Alan,


"Alan Smallbone" <asmallbone@...> writes:
> in my program I have the following setup for my file writing:
> #define mywrite(fspec, count, data) count=fwrite(data,1,count,fspec)
>
> For discussion lets say count=4 and data is pointing to "ABCD"
>
> When I compile under Mingw ver1.08 and write out the data I get
> "ABCD" in the file which is what I want. When compiled with the
> latest release of Mingw 2.0. and executed I get "DCBA" in the file,

The behaviour of the data sounds like a byte-order problem but your
description doesn't.  Can you provide a 10-20 lines compilable program
to illustrate the problem that you have?

Certainly the compiler hasn't changed in the byte-ordering aspect, not
is it likely that there are or ever will be switches to change that
type of behaviour.


so long, benny

Re: [Mingw-users] Fw: A DLL problem

[Roger K. Wells <ROGER.K.WELLS@sa...>]

Virtual wrote:

> I have a little problem sharing variables from a DLL to 
> another...please see this example:
>  
> -------- FILE 1: hello.c --------
>  
> #include "dllfct.h"
>  
> int main () {
>   i = 5;
>   tstfunc();
>   i = 2;
>   tstfunc2();
>   return (0);
> }
>  
> --------- FILE 2: dllfct.h -------
>  
> #ifdef BUILD_DLL
>   #define EXPORT __declspec(dllexport)
> #else
>   #define EXPORT __declspec(dllimport)
> #endif
>  
> EXPORT int i;
> EXPORT void tstfunc (void);
>  
> --------- FILE 3: dllfct.c ----------
>  
>    #include <stdio.h>
>    #include "dllfct.h"
>  
>    EXPORT int i=0;
>  
>    EXPORT void tstfunc (void) {
>       printf ("Hello%d\n", i);
>    }
> ---------- FILE 4: dllfct2.h -----------
> EXPORT void tstfunc2 (void);
> ---------- FILE 5: dllfct2.c -----------
>  
> #include <stdio.h>
> #include "dllfct.h"
> #include "dllfct2.h"
>  
> EXPORT void tstfunc2 (void) {
>   printf("VAL IS=%d", i);
> }
>  
> ---------- COMPILED WITH: --------------
>  
> gcc -DBUILD_DLL -shared -o tst.dll  dllfct.c 
> -Wl,--output-def,tstdll.def,--out-implib,libtstdll.a
>  
> gcc -DBUILD_DLL -shared -o tst2.dll  dllfct2.c 
> -Wl,--output-def,tst2dll.def,--out-implib,libtstdll2.a
>  
> gcc -o hello.exe hello.o -L./ -ltstdll -ltstdll2
>  
> ----------- BUT ON VIDEO: ----------------
>  
> Hello5
> VAL IS=0
>  
> --------------------------------------------
>  
> how can I view the changes to i, maded by  hello.c, from tstfunc2 into 
> tst2.dll ????
>  
> Please, it's very important.
>  
>  
> P.S: 'scuse for my bad english!!!

you need two BUILD_DLL macros.  This because when you are building 
dllfct2.dll you are in fact using dllfct.dll.  Define DLLFCT_BUILD_DLL & 
DLLFCT2_BUILD_DLL & define them as appropriate

hope this helps

-- 
Roger Wells, P.E.
SAIC
221 Third St
Newport, RI 02840
401-847-4210 (voice)
401-849-1585 (fax)
roger@...

[Mingw-users] gcc 323 sjlj exception handling model

[Serge Bakkal <Serge.Bakkal@wa...>]

Hi , this is my first post to list so I hope I formulate this well :
I'm compiling a large framework using mingw while the original compiler on
nt stations is msvc .

I' ve been experimenting troubles with exception handling :
gcc323 (2003-05-04) compiles with option --enable-sjlj-exceptions (reading
3.2.3/specs )
but it appears that __USING_SJLJ_EXCEPTIONS__ has to be defined to get the
correct handler
 linked with excutables (dll ,exe) .

Is that expected behavior , and if not , is this the relevant list where
this should be posted ?

Serge Bakkal

[Mingw-users] Fw: A DLL problem

[Virtual <virtual-heaven@li...>]

I have a little problem sharing variables from a DLL to another...please =
see this example:

-------- FILE 1: hello.c --------

#include "dllfct.h"

int main () {
  i =3D 5;
  tstfunc();
  i =3D 2;
  tstfunc2();
  return (0);
}

--------- FILE 2: dllfct.h -------

#ifdef BUILD_DLL
  #define EXPORT __declspec(dllexport)
#else
  #define EXPORT __declspec(dllimport)
#endif

EXPORT int i;
EXPORT void tstfunc (void);

--------- FILE 3: dllfct.c ----------

   #include <stdio.h>
   #include "dllfct.h"

   EXPORT int i=3D0;

   EXPORT void tstfunc (void) {
      printf ("Hello%d\n", i);
   }

---------- FILE 4: dllfct2.h -----------
EXPORT void tstfunc2 (void);
---------- FILE 5: dllfct2.c -----------

#include <stdio.h>
#include "dllfct.h"
#include "dllfct2.h"

EXPORT void tstfunc2 (void) {
  printf("VAL IS=3D%d", i);
}

---------- COMPILED WITH: --------------

gcc -DBUILD_DLL -shared -o tst.dll  dllfct.c =
-Wl,--output-def,tstdll.def,--out-implib,libtstdll.a

gcc -DBUILD_DLL -shared -o tst2.dll  dllfct2.c =
-Wl,--output-def,tst2dll.def,--out-implib,libtstdll2.a

gcc -o hello.exe hello.o -L./ -ltstdll -ltstdll2

----------- BUT ON VIDEO: ----------------

Hello5
VAL IS=3D0

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

how can I view the changes to i, maded by  hello.c, from tstfunc2 into =
tst2.dll ????

Please, it's very important.


P.S: 'scuse for my bad english!!!

Re: [Mingw-users] Program Distribution - Installers (win32/unix)

[Greg Chicares <chicares@mi...>]

Peter Mcleod wrote:
> 
[...]
> Whatever info you have, recommendations or sites that you
> can point me to would be appreciated

My favorite installer is tar + bzip2, so I don't have
any positive recommendation for GUI installers--but I
do have a negative one: avoid any 16-bit installer.

Some 32-bit windows packages are still distributed with 
16-bit install programs. Maybe that's for backward
compatibility with win32s; I dunno. But to run these
installers under windows 2K, they need NTVDM. Yet NTVDM
is unsafe and can crash the OS, so I've removed it from
my system (I'm not the only one who's done this), and
therefore 16-bit installers won't run here.

Re: [Mingw-users] Compiler changes

[Roger K. Wells <ROGER.K.WELLS@sa...>]

Alan Smallbone wrote:

>Hi,
>
>in my program I have the following setup for my file writing:
>#define mywrite(fspec, count, data) count=fwrite(data,1,count,fspec)
>
>For discussion lets say count=4 and data is pointing to "ABCD"
>
>When I compile under Mingw ver1.08 and write out the data I get "ABCD" in
>the file which is what I want. When compiled with the latest release of
>Mingw 2.0. and executed I get "DCBA" in the file, do I need to rewrite my
>program or is there a switch that will change this to behave as it did in
>the previous release? I am interested in using the latest version if
>possible without rewriting my program. Or there is probably a glaring error
>in how I am implementing this, any thoughts would be appreciated. If this is
>the wrong forum please let me know.
>
>Thanks in advance for any help.
>
>Alan
>
>
>Alan Smallbone
>FEM, Inc.
>FAA Repair Station PU3R776L
>(818) 240-7740
>(818) 240-1009 fax
>asmallbone@...
>http://www.fortnereng.com
>
>
>
>
>-------------------------------------------------------
>This SF.net email is sponsored by: ObjectStore.
>If flattening out C++ or Java code to make your application fit in a
>relational database is painful, don't do it! Check out ObjectStore.
>Now part of Progress Software. http://www.objectstore.net/sourceforge
>_______________________________________________
>MinGW-users mailing list
>MinGW-users@...
>
>You may change your MinGW Account Options or unsubscribe at:
>https://lists.sourceforge.net/lists/listinfo/mingw-users
>
>
>  
>
FWIW the following works as expected W2k, gcc (GCC) 3.2.3 (mingw special 
20030425-1):

#include <stdio.h>
#include <stdlib.h>

#define mywrite(fspec, count, data) count=fwrite(data,1,count,fspec)

int main(void)
{
    FILE *f;
    char *data = "ABCD";
    int count = 4;

    f = fopen("fwrite.out", "wb+");
    if (f) {
        mywrite(f, count, data);
        fclose(f);
    }


    return 0;
} 

od -t x1 -a fwrite.out yields:
0000000 41 42 43 44
          A   B   C   D
0000004

for at least setting file mode to: "wb", "wb+", "w+", "w", "r+" (if it 
exists), "a+"


-- 
Roger Wells, P.E.
SAIC
221 Third St
Newport, RI 02840
401-847-4210 (voice)
401-849-1585 (fax)
roger@...

Re: [Mingw-users] Program Distribution - Installers (win32/unix)

[Earnie Boyd <earnie_boyd@ya...>]

Peter Mcleod wrote:
> Hi all,
> 
> When it comes to installation programs, a preference I have
> always had is for a friendly self-extracting (compressed)
> executable that automatically copies its contents to a temp
> directory and then starts up an install program (this is
> often helpful as there are still many computer users out
> there who are not so literate).  The MinGW and MSYS
> binaries themselves are provided in such a download format.
>  

Thanks, I agree.

> 
> My questions are:
> 
> 1.  Ideally, I would like to package my cross-os
> programs/utilities like this - is there anything in the
> minGW/gcc world for setting up something similar so that it
> produces a win32 and a unix installer?  How did they
> achieve it with the MinGW installer?  What was used for the
> GUI - surely not GTK without a runtime?  --It may help to
> know I always keep my source code 100% compatible between
> win32 and unix - that is, no macros, no using os specific
>  function/run-time libraries.  I don't even write to the
> windows registery but expect that the installer would have
> to - for the windows "add/remove programs" utility.
> 

I use an open source utility named INNO Setup which you can find at 
http://www.jrsoftware.org/isinfo.php and a bit of bash scripting.  The 
drawback is that the source is Borland Pascal.  INNO Setup uses a script 
to control what is included and has pre and post install controls.  It 
doesn't have to write to registry at all but will if you choose to 
include uninstall procedures only for the "add/remove programs" utility.

> 2.  My previous experience with install programs are all
> win32 related - I understand the various linux
> distributions have their own package managers (I've used
> RPM quite a bit but that's all).  How does the linux world
> co-ordinate installations (if at all)?  I wouldn't want to
> lock out anybody from using any of my programs/utilities
> because they don't install anything that doesn't go through
> their package manager but at the same time I don't want to
> create and maintain heaps and heaps of install packages?  
> 

What a delima, I would hope that package distributors using RPM and 
DEBIAN dpkg have some nice utilities to help that delima.  I don't know 
if you can get the INNO Setup to build for linux, I doubt it, it's 
mostly Win32 oriented.  I do know that there are ports of some style of 
both RPM and DPKG for win32, there is even a debian-win32 list at 
debian.org.

> Whatever info you have, recommendations or sites that you
> can point me to would be appreciated?  If I have to pay for
> a utility (that supports win32 and linux) and it saves me
> the bother of too much maintenance, I would be glad to.  
> 

Well, let me know what you come up with.  I certainly don't have an 
answer for something Linux.  INNO Setup works well for Win32.

> I have yet another question (which I hope won't detract
> anybody from the above and more important to me) - and that
> is, how much like linux/unix is mac-osx - do we have a gcc
> compiler available for it - the standard compiler?  If I
> write a plain ansi c program in linux and compile it will
> it run on linux - or do I need to recompile it on that os?
>  

IIRC, gcc does exist for OSX.  You will need to recompile and link your 
binaries unless emulation libraries are provided, such as WINE or LINE.

HTH,
Earnie.

[Mingw-users] Program Distribution - Installers (win32/unix)

[Peter Mcleod <peter_m@an...>]

Hi all,

When it comes to installation programs, a preference I have
always had is for a friendly self-extracting (compressed)
executable that automatically copies its contents to a temp
directory and then starts up an install program (this is
often helpful as there are still many computer users out
there who are not so literate).  The MinGW and MSYS
binaries themselves are provided in such a download format.
 

My questions are:

1.  Ideally, I would like to package my cross-os
programs/utilities like this - is there anything in the
minGW/gcc world for setting up something similar so that it
produces a win32 and a unix installer?  How did they
achieve it with the MinGW installer?  What was used for the
GUI - surely not GTK without a runtime?  --It may help to
know I always keep my source code 100% compatible between
win32 and unix - that is, no macros, no using os specific
 function/run-time libraries.  I don't even write to the
windows registery but expect that the installer would have
to - for the windows "add/remove programs" utility.

2.  My previous experience with install programs are all
win32 related - I understand the various linux
distributions have their own package managers (I've used
RPM quite a bit but that's all).  How does the linux world
co-ordinate installations (if at all)?  I wouldn't want to
lock out anybody from using any of my programs/utilities
because they don't install anything that doesn't go through
their package manager but at the same time I don't want to
create and maintain heaps and heaps of install packages?  

Whatever info you have, recommendations or sites that you
can point me to would be appreciated?  If I have to pay for
a utility (that supports win32 and linux) and it saves me
the bother of too much maintenance, I would be glad to.  

I have yet another question (which I hope won't detract
anybody from the above and more important to me) - and that
is, how much like linux/unix is mac-osx - do we have a gcc
compiler available for it - the standard compiler?  If I
write a plain ansi c program in linux and compile it will
it run on linux - or do I need to recompile it on that os?
 


Thanks,


Peter.
 
==
Download ringtones, logos and picture messages at Ananzi Mobile Fun.
http://www.ananzi.co.za/cgi-bin/goto.pl?mobile

Re: [Mingw-users] Compiler changes

[Oscar Fuentes <ofv@wa...>]

"Alan Smallbone" <asmallbone@...> writes:

> in my program I have the following setup for my file writing:
> #define mywrite(fspec, count, data) count=fwrite(data,1,count,fspec)
>
> For discussion lets say count=4 and data is pointing to "ABCD"
>
> When I compile under Mingw ver1.08 and write out the data I get "ABCD" in
> the file which is what I want. When compiled with the latest release of
> Mingw 2.0. and executed I get "DCBA" in the file,
[snip]

That's *very* strange. Could you provide a test case?

-- 
Oscar