RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Christopher Jones]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 3022 bytes --]

Title: RE: [mingw32] Re: [RFC] changing gcc default output executable na me         (a.exe now)





As far as I know the problems which were solved by having the .exe automatically appended to the executable's name could have just as easily been dealt with via Makefiles, and I agree it probably creates more problems than it solves.  For that matter why couldn't we have just been forced to do something like gcc -o a.exe a.c instead?  This interoperability tradeoff has a cost in terms of emulation.

Related to another thread, I think cygwin goes after two slightly different objectives at the same time and it creates problems when you talk about things like the POSIX filenames mentioned earlier.  On the one hand you want to have a UNIX emulation layer and on the other you want to have interoperability with the Windows environment.  I personally place a lot less emphasis on the latter.  Windows be damned.  Maybe there should be two environments, one for compromises to induce interoperability and another without those compromises.  My guess is that the latter would be easier to setup and easier to use.

Brian


> -----Original Message-----
> From: Andre Oliveira da Costa [ mailto:costa@cade.com.br ]
> Sent: Thursday, January 13, 2000 2:30 PM
> To: cygwin@sourceware.cygnus.com
> Subject: RE: [mingw32] Re: [RFC] changing gcc default output 
> executable
> na me (a.exe now)
> 
> 
> Mmmh... ok, I agree this would solve the problem for packages 
> which use
> autoconf/automake. But those who don't are still in the dark. 
> The worst of
> it all is that the .exe suffix is added by ld -- it is as if 
> cygwin creates
> problems for itself...
> 
> (before flaming starts: I know there's a very good reason for 
> the insertion
> of the .exe suffix: it has to be added so that the files can 
> be seen as
> executables by Windoze -- another great feature (?) by 
> M$oft... =T As I
> said, I just think it would be nice to have this .exe suffix handled
> transparently, since it is automagically created)
> 
> Regards,
> 
> Andre
> --
> André Oliveira da Costa
> (costa@cade.com.br)
> 
> -----Original Message-----
> From: cygwin-owner@sourceware.cygnus.com
> [ mailto:cygwin-owner@sourceware.cygnus.com ]On Behalf Of 
> Christopher Jones
> Sent: Thursday, January 13, 2000 5:03 PM
> To: Andre Oliveira da Costa; cygwin@sourceware.cygnus.com
> Subject: RE: [mingw32] Re: [RFC] changing gcc default output 
> executable na
> me (a.exe now)
> 
> I think this should be handled automatically if using 
> autoconf and automake,
> especially when using automake since the install targets for 
> executables are
> created by it from bin_PROGS or similar.  The cygwin stuff in autoconf
> provides a variable which can be checked in Makefiles to do something
> specific for a cygwin environment and another variable to do something
> cygwin specific in configure.
> Brian
> 
> 
> --
> Want to unsubscribe from this list?
> Send a message to cygwin-unsubscribe@sourceware.cygnus.com
> 

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Andre Oliveira da Costa]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1683 bytes --]

> -----Original Message-----
> From: cygwin-owner@sourceware.cygnus.com
> [ mailto:cygwin-owner@sourceware.cygnus.com]On Behalf Of Chris Faylor
> Sent: Friday, January 14, 2000 5:07 PM
> To: Andre Oliveira da Costa
> Cc: rbh00@netcom.com; cygwin@sourceware.cygnus.com
> Subject: Re: [mingw32] Re: [RFC] changing gcc default output executable
> na me (a.exe now)
>
> We were talking about having gcc understand an environment variable or
> command line switch causing it to avoid producing a .exe extension.  It
> is not going to be the default.  Nothing is going to change unless you
> set this.

Looks fine to me, I can live with that ;) Let the default be the current
behavior, and people who don't need/want .exe suffixes automatically
appended to executables (like me) can have some variable set on their
.bashrc files that would turn off this automatic feature.

> Then, of course, the mailing list will be flooded with lots of "newbies"
> who have found out about this nifty new environment variable and who
> 1) have an older version of gcc which doesn't understand it, 2)
> have makefiles
> which add the .exe file explicitly, or 3) are facing in the wrong
> direction
> and are unable to see the computer screen.

[1] they would be on the same situation we are today: they'd like not to
have the .exe appended, but would have to tweak makefiles for that

[2] these should work with no changes, since they are supposedly able to
handle executables with the .exe suffix

[3] these should be killed. Slowly ;)

Regards,

Andre
--
André Oliveira da Costa
(costa@cade.com.br)


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Chris Faylor]

On Fri, Jan 14, 2000 at 02:41:40PM -0200, Andre Oliveira da Costa wrote:
>> If such a patch as this goes forward, am I right that the implicit
>> make(1) rules will be changed at the same time gcc and/or binutils is
>> changed?  I mean that
>> 	make hello
>> would by default produce the file "hello" rather than "hello.exe".
>
>Yep. Going a step further, every rule that is meant to produce an executable
>file, like:
>
>foo: $(OBJS)
>	gcc -o $@ $^
>
>would produce an executable (binary) file named "foo", and not "foo.exe"

We were talking about having gcc understand an environment variable or
command line switch causing it to avoid producing a .exe extension.  It
is not going to be the default.  Nothing is going to change unless you
set this.

Then, of course, the mailing list will be flooded with lots of "newbies"
who have found out about this nifty new environment variable and who
1) have an older version of gcc which doesn't understand it, 2) have makefiles
which add the .exe file explicitly, or 3) are facing in the wrong direction
and are unable to see the computer screen.


cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Andre Oliveira da Costa]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1718 bytes --]

> -----Original Message-----
> From: Richard Hitt [ mailto:rbh00@netcom.com ]
> Sent: Friday, January 14, 2000 1:55 AM
> To: Andre Oliveira da Costa
> Cc: cygwin@sourceware.cygnus.com
> Subject: Re: [mingw32] Re: [RFC] changing gcc default output executable
> na me (a.exe now)
>
> Hi all
>
> If such a patch as this goes forward, am I right that the implicit
> make(1) rules will be changed at the same time gcc and/or binutils is
> changed?  I mean that
> 	make hello
> would by default produce the file "hello" rather than "hello.exe".

Yep. Going a step further, every rule that is meant to produce an executable
file, like:

foo: $(OBJS)
	gcc -o $@ $^

would produce an executable (binary) file named "foo", and not "foo.exe"
like it happens today with the actual gcc / ld tools. This behavior
(appending ".exe" on the output file name) happens only on Windows, because
the suffix is necessary so that Windows can see the file as an executable
(another braindamage(TM) by Micro$oft). This creates a problem for UNIX
makefiles, because none of the rules consider a .exe suffix when they refer
to the executable.

> And if not, why not?  :-)

Ask Micro$oft... the .exe suffix is needed only if you have to run your
application through Windows tools (cmd.exe or Windows Explorer, for
example). If you only use your cygwin apps through bash prompt or other
cygwin apps, there's no need for the suffix. What we're proposing here is
that this suffix should only be appended when it is really needed, and this
should be controlled by environment variables.

HTH,

Andre
--
André Oliveira da Costa
(costa@cade.com.br)


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Richard Hitt]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1190 bytes --]

Hi all

If such a patch as this goes forward, am I right that the implicit
make(1) rules will be changed at the same time gcc and/or binutils is
changed?  I mean that
	make hello
would by default produce the file "hello" rather than "hello.exe".

And if not, why not?  :-)

Richard




On Thu, 13 Jan 2000 19:37:22 -0200, you wrote:

>> If someone wants to submit a patch to the gcc maintainers I'm sure that
>> they will give it their attention.
>>
>> In fact, if you Cc me, I will talk to people at Cygnus about it and try
>> to advocate that it be installed.
>>
>> I would suggest that somebody who feels strongly about it should start
>> working on this ASAP.
>
>That's good. Any volunteers? I'll try to take a look at it ASAP (which might
>take sometime... please, if anybody else has the chance, feel free to do it
>first). But, who is the actual responsible for appending the .exe suffix,
>gcc or ld? Also, I'd like to propose a change from GCC_SUFFIX to EXE_SUFFIX.
>What do you (Chris, Earnie, cygwin-community) think?
>
>Andre
>--
>André Oliveira da Costa
>(costa@cade.com.br)


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Andre Oliveira da Costa]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 837 bytes --]

> If someone wants to submit a patch to the gcc maintainers I'm sure that
> they will give it their attention.
>
> In fact, if you Cc me, I will talk to people at Cygnus about it and try
> to advocate that it be installed.
>
> I would suggest that somebody who feels strongly about it should start
> working on this ASAP.

That's good. Any volunteers? I'll try to take a look at it ASAP (which might
take sometime... please, if anybody else has the chance, feel free to do it
first). But, who is the actual responsible for appending the .exe suffix,
gcc or ld? Also, I'd like to propose a change from GCC_SUFFIX to EXE_SUFFIX.
What do you (Chris, Earnie, cygwin-community) think?

Andre
--
André Oliveira da Costa
(costa@cade.com.br)


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Chris Faylor]

On Thu, Jan 13, 2000 at 06:55:27PM -0200, Andre Oliveira da Costa wrote:
>>I agree, Andre.  I suggested, years ago now, that the .exe suffix be an
>>optional configuration whose default is off.  It isn't needed by the
>>Bourne shells we use to execute the programs.  I would like to see this
>>as optional with say a GCC_SUFFIX=.exe environment variable.  If
>>GCC_SUFFIX exists it uses the value to append to the executable output
>>file.
>
>It seems like a good (and simple) solution to me.  With that, UNIX
>software compiled directly from source would compile/install flawlessly
>if GCC_SUFFIX is undefined or defined as an empty var.  And, the folks
>who need the .exe suffix (as stated by Matthew on a message to this
>thread) could just add "export GCC_SUFFIX=" to their .bashrc startup
>files.

If someone wants to submit a patch to the gcc maintainers I'm sure that
they will give it their attention.

In fact, if you Cc me, I will talk to people at Cygnus about it and try
to advocate that it be installed.

I would suggest that somebody who feels strongly about it should start
working on this ASAP.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Andre Oliveira da Costa]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1223 bytes --]

> -----Original Message-----
> From: cygwin-owner@sourceware.cygnus.com
> [ mailto:cygwin-owner@sourceware.cygnus.com]On Behalf Of Earnie Boyd
> Sent: Thursday, January 13, 2000 6:27 PM
> To: Andre Oliveira da Costa; cygwin@sourceware.cygnus.com
> Subject: RE: [mingw32] Re: [RFC] changing gcc default output executable
> na me (a.exe now)
>
> I agree, Andre.  I suggested, years ago now, that the .exe suffix be an
> optional configuration whose default is off.  It isn't needed by
> the Bourne
> shells we use to execute the programs.  I would like to see this
> as optional
> with say a GCC_SUFFIX=.exe environment variable.  If GCC_SUFFIX
> exists it uses
> the value to append to the executable output file.

It seems like a good (and simple) solution to me. With that, UNIX software
compiled directly from source would compile/install flawlessly if GCC_SUFFIX
is undefined or defined as an empty var. And, the folks who need the .exe
suffix (as stated by Matthew on a message to this thread) could just add
"export GCC_SUFFIX=" to their .bashrc startup files.

Andre
--
André Oliveira da Costa
(costa@cade.com.br)


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Matthew Brown]

I could see allowing an option to drop the extension; however, most of us who do cross-platform stuff have customers on Windows who do not use shells at all. For this reason along with that thing about backward-compatibility, I would only want to see it as an option and not the default behavior.

-- Matthew Brown

----- Original Message ----- 
From: "Earnie Boyd" <earnie_boyd@yahoo.com>
To: "Andre Oliveira da Costa" <costa@cade.com.br>; <cygwin@sourceware.cygnus.com>
Sent: Thursday, January 13, 2000 3:26 PM
Subject: RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)


> --- Andre Oliveira da Costa <costa@cade.com.br> wrote:
> > Mmmh... ok, I agree this would solve the problem for packages which use
> > autoconf/automake. But those who don't are still in the dark. The worst of
> > it all is that the .exe suffix is added by ld -- it is as if cygwin creates
> > problems for itself...
> > 
> > (before flaming starts: I know there's a very good reason for the insertion
> > of the .exe suffix: it has to be added so that the files can be seen as
> > executables by Windoze -- another great feature (?) by M$oft... =T As I
> > said, I just think it would be nice to have this .exe suffix handled
> > transparently, since it is automagically created)
> > 
> 
> I agree, Andre.  I suggested, years ago now, that the .exe suffix be an
> optional configuration whose default is off.  It isn't needed by the Bourne
> shells we use to execute the programs.  I would like to see this as optional
> with say a GCC_SUFFIX=.exe environment variable.  If GCC_SUFFIX exists it uses
> the value to append to the executable output file.
> 
> 
> =====
> Earnie Boyd < mailto:earnie_boyd@yahoo.com >
> Cygwin Newbies, please visit
> < http://www.freeyellow.com/members5/gw32/index.html >
> __________________________________________________
> Do You Yahoo!?
> Talk to your friends online with Yahoo! Messenger.
> http://im.yahoo.com
> 
> --
> Want to unsubscribe from this list?
> Send a message to cygwin-unsubscribe@sourceware.cygnus.com
> 
> 


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Andre Oliveira da Costa]

> -----Original Message-----
> From: cygwin-owner@sourceware.cygnus.com
> [ mailto:cygwin-owner@sourceware.cygnus.com]On Behalf Of Chris Faylor
> Sent: Thursday, January 13, 2000 6:08 PM
> To: Andre Oliveira da Costa
> Cc: cygwin@sourceware.cygnus.com
> Subject: Re: [mingw32] Re: [RFC] changing gcc default output executable
>
> As usual, my suggestion is that rather than opine how something should be
> handled "automatically", actual source code demonstrating your idea would
> be much more helpful.
>
> The sources to cp, cygwin, mv, rm, make, gcc, etc.  are all available.
> Personally, I have no idea how you could implement a plan to
> "automatically" do the right thing in such a way that you will end up
> satisfying everybody.  However, I'd love to be proved wrong by seeing
> exactly what you are proposing.
>
> cgf

I didn't mean I had an alternative in mind, and I was not suggesting a
specific course of action. I just wanted to start some discussion to see if
some ideas were proposed.

What I said about the file tools (cp, mv, install etc.) is that if they were
able to understand that a missing "foo" file shouldn't be treated as an
error if there was a "foo.exe" file would solve the problem as for their use
in makefiles. But, it's clearly not a good idea (and I said that on my
email) because you have many drawbacks such as loss of control (this
aliasing foo = foo.exe would probably create more problems than solutions)
and performance (there would be an overhead of extra checks for missing
files situations).

Don't get me wrong: I love cygwin, and I think you guys do an amazing work.
I use it a lot at work, and it's extremely useful to me. The only purpose of
this discussion is to try to make it an even better tool, not to create
problems for no reason.

Best regards,

Andre
--


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Earnie Boyd]

--- Andre Oliveira da Costa <costa@cade.com.br> wrote:
> Mmmh... ok, I agree this would solve the problem for packages which use
> autoconf/automake. But those who don't are still in the dark. The worst of
> it all is that the .exe suffix is added by ld -- it is as if cygwin creates
> problems for itself...
> 
> (before flaming starts: I know there's a very good reason for the insertion
> of the .exe suffix: it has to be added so that the files can be seen as
> executables by Windoze -- another great feature (?) by M$oft... =T As I
> said, I just think it would be nice to have this .exe suffix handled
> transparently, since it is automagically created)
> 

I agree, Andre.  I suggested, years ago now, that the .exe suffix be an
optional configuration whose default is off.  It isn't needed by the Bourne
shells we use to execute the programs.  I would like to see this as optional
with say a GCC_SUFFIX=.exe environment variable.  If GCC_SUFFIX exists it uses
the value to append to the executable output file.


=====
Earnie Boyd < mailto:earnie_boyd@yahoo.com >
Cygwin Newbies, please visit
< http://www.freeyellow.com/members5/gw32/index.html >
__________________________________________________
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.
http://im.yahoo.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Chris Faylor]

On Thu, Jan 13, 2000 at 05:30:03PM -0200, Andre Oliveira da Costa wrote:
>Mmmh... ok, I agree this would solve the problem for packages which use
>autoconf/automake. But those who don't are still in the dark. The worst of
>it all is that the .exe suffix is added by ld -- it is as if cygwin creates
>problems for itself...
>
>(before flaming starts: I know there's a very good reason for the insertion
>of the .exe suffix: it has to be added so that the files can be seen as
>executables by Windoze -- another great feature (?) by M$oft... =T As I
>said, I just think it would be nice to have this .exe suffix handled
>transparently, since it is automagically created)

As usual, my suggestion is that rather than opine how something should be
handled "automatically", actual source code demonstrating your idea would
be much more helpful.

The sources to cp, cygwin, mv, rm, make, gcc, etc.  are all available.
Personally, I have no idea how you could implement a plan to
"automatically" do the right thing in such a way that you will end up
satisfying everybody.  However, I'd love to be proved wrong by seeing
exactly what you are proposing.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Andre Oliveira da Costa]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1489 bytes --]

Mmmh... ok, I agree this would solve the problem for packages which use
autoconf/automake. But those who don't are still in the dark. The worst of
it all is that the .exe suffix is added by ld -- it is as if cygwin creates
problems for itself...

(before flaming starts: I know there's a very good reason for the insertion
of the .exe suffix: it has to be added so that the files can be seen as
executables by Windoze -- another great feature (?) by M$oft... =T As I
said, I just think it would be nice to have this .exe suffix handled
transparently, since it is automagically created)

Regards,

Andre
--
André Oliveira da Costa
(costa@cade.com.br)

-----Original Message-----
From: cygwin-owner@sourceware.cygnus.com
[ mailto:cygwin-owner@sourceware.cygnus.com]On Behalf Of Christopher Jones
Sent: Thursday, January 13, 2000 5:03 PM
To: Andre Oliveira da Costa; cygwin@sourceware.cygnus.com
Subject: RE: [mingw32] Re: [RFC] changing gcc default output executable na
me (a.exe now)

I think this should be handled automatically if using autoconf and automake,
especially when using automake since the install targets for executables are
created by it from bin_PROGS or similar.  The cygwin stuff in autoconf
provides a variable which can be checked in Makefiles to do something
specific for a cygwin environment and another variable to do something
cygwin specific in configure.
Brian


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: [mingw32] Re: [RFC] changing gcc default output executable na me (a.exe now)

[Christopher Jones]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 838 bytes --]

Title: RE: [mingw32] Re: [RFC] changing gcc default output executable name (a.exe now)





> I'd like to ask for comments on another issue related to 
> portability from
> UNIX --> cygwin, compilation etc.: the way it is now, the linker
> automagically appends a .exe suffix to the executable 
> filename. If you do
> 
> gcc -o foo foo.o
> 
> ld will create foo.exe .
> ...
> Am I missing something or is this a real problem?


I think this should be handled automatically if using autoconf and automake, especially when using automake since the install targets for executables are created by it from bin_PROGS or similar.  The cygwin stuff in autoconf provides a variable which can be checked in Makefiles to do something specific for a cygwin environment and another variable to do something cygwin specific in configure.

Brian