Re: Make and javac compliler problem in bash

Re: Make and javac compliler problem in bash

[Woody Jin]

At 11:53 AM 6/12/00 -0400, you wrote:
>At 10:21 AM 6/12/00 -0700, you wrote:
> ><Problem 1>
> >I have a Makefile which compiles java program, if I type
> >make, after the screen shows the "javac xxxx.java", and then
> >hangs.  The Makefile doesn't need to
> >be complex. Any simple Makefile that compiles even HelloWorld.java
> >will make the bash console stuck.
>
>See "Why is make behaving badly?" in the Cygnus faq at
> http://sourceware.cygnus.com/cygwin/faq/

1) I think that you are misunderstanding the problem.
2) You think that everything is in FAQ ?
Whatever you do, "make" will stuck with javac.


> ><Problem 2>
> >It seems to me that javac in Windows will accept directories only in
> >MS-DOS way. For example, "javac -classpath /a/b/c" won't work.
> >
>
>How could it be otherwise? The Windows version of javac is, surprise, a
>Windows application. Windows applications don't recognize Unix paths. Ergo,
>the Windows version of javac does not recognize Unix paths. QED.


Why can't it be otherwise ? Lack of imagination ?  QED ( :-) )
1) I thought that javac, being a product of Sun Microsystem, (and there exist
     its counterpart in Unix) would accept unix style path, which I found 
is  not
     the case. It is better to write an application to accept both styles. 
Maybe
     we should request Sun to do it, letting them know that in Windows there
     are many Unix tool users.
2) Rather than messing around the backslash and slash all over the places,
     and make Makefiles and other shell scripts incompatible with the Unix,
     (for example, I want to use the same Makefile on both platform - yes, 
I have
      Sunworkstation on my left side and WindowsNT on my right side),
     maybe, bash or some other utility registers applications which requires
     windows specific path, and whenever you use unix style path, it 
automatically
     converts to MS-DOS style when the system finally gives the application 
the path.





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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 01:26 PM 6/12/00 -0700, Woody Jin wrote:
>At 11:53 AM 6/12/00 -0400, you wrote:
>>At 10:21 AM 6/12/00 -0700, you wrote:
>> ><Problem 1>
>> >I have a Makefile which compiles java program, if I type
>> >make, after the screen shows the "javac xxxx.java", and then
>> >hangs.  The Makefile doesn't need to
>> >be complex. Any simple Makefile that compiles even HelloWorld.java
>> >will make the bash console stuck.
>>
>>See "Why is make behaving badly?" in the Cygnus faq at
>> http://sourceware.cygnus.com/cygwin/faq/
>
>1) I think that you are misunderstanding the problem.
>2) You think that everything is in FAQ ?
>Whatever you do, "make" will stuck with javac.
>

Did you actually try running make in Unix mode as suggested in the FAQ?

>
>> ><Problem 2>
>> >It seems to me that javac in Windows will accept directories only in
>> >MS-DOS way. For example, "javac -classpath /a/b/c" won't work.
>> >
>>
>>How could it be otherwise? The Windows version of javac is, surprise, a
>>Windows application. Windows applications don't recognize Unix paths. Ergo,
>>the Windows version of javac does not recognize Unix paths. QED.
>
>
>Why can't it be otherwise ? Lack of imagination ?  QED ( :-) )
>1) I thought that javac, being a product of Sun Microsystem, (and there exist
>     its counterpart in Unix) would accept unix style path, which I found 
>is  not
>     the case. It is better to write an application to accept both styles. 
>Maybe
>     we should request Sun to do it, letting them know that in Windows there
>     are many Unix tool users.
>2) Rather than messing around the backslash and slash all over the places,
>     and make Makefiles and other shell scripts incompatible with the Unix,
>     (for example, I want to use the same Makefile on both platform - yes, 
>I have
>      Sunworkstation on my left side and WindowsNT on my right side),
>     maybe, bash or some other utility registers applications which requires
>     windows specific path, and whenever you use unix style path, it 
>automatically
>     converts to MS-DOS style when the system finally gives the application 
>the path.
>

Think about what you're asking. If you come up with a general, complete
solution to the problem of unambiguously converting DOS to Unix paths, and
vice-versa, without any additional information than what is in the paths
(don't forget the little problem of drive letters), patent it. A lot of
people would be very interested in your solution.

- Paul


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

Re: Make and javac compliler problem in bash

[Jason Tishler]

Jin,

Woody Jin wrote:
> 1) I thought that javac, being a product of Sun Microsystem, (and there exist
>      its counterpart in Unix) would accept unix style path, which I found
> is  not
>      the case.

Sun's javac does support paths with forward slashes -- but will not
handle Cygwin style path's (via mounts, symlinks, etc.).

> 2) Rather than messing around the backslash and slash all over the places,
>      and make Makefiles and other shell scripts incompatible with the Unix,
>      (for example, I want to use the same Makefile on both platform

We successfully use GNU make (ie, 3.79) and javac on both NT and UNIX
without any problems.  See attached for an example.

Sorry, that I can't give you specific help, but hopefully knowing that
it should work will help you persevere until you succeed.

Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com
# $Id: Makefile,v 1.1 2000/05/05 13:58:58 jt Exp $

BaseDir = ../..

include $(BaseDir)/Makerules.mk

all:: main

main: $(CLASSES)
	$(JAVAC) $(JAVACFLAGS) *.java \
		../stgobjs/aryobjs/*.java ../stgobjs/spobjs/*.java ../stgobjs/*.java \
		../server/*.java \
		../ui/cfg_ui/*.java ../ui/*.java

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 03:21 PM 6/12/00 -0400, Jason Tishler wrote:
>Jin,
>
>Woody Jin wrote:
>> 1) I thought that javac, being a product of Sun Microsystem, (and there
exist
>>      its counterpart in Unix) would accept unix style path, which I found
>> is  not
>>      the case.
>
>Sun's javac does support paths with forward slashes -- but will not
>handle Cygwin style path's (via mounts, symlinks, etc.).
>

The Windows version of javac does not support paths that BEGIN with a
forward slash. That is the crux of Jin's complaint. 

>> 2) Rather than messing around the backslash and slash all over the places,
>>      and make Makefiles and other shell scripts incompatible with the Unix,
>>      (for example, I want to use the same Makefile on both platform
>
>We successfully use GNU make (ie, 3.79) and javac on both NT and UNIX
>without any problems.  See attached for an example.
>

Yes, but have you run the Cygnus version of GNU make on Windows/NT?  The
Cygnus version of GNU make is a version that has been modified to run in
two modes: Windows mode and  Unix mode. See the Cygnus FAQ for more
information.  I've generally had the most success running the Cygnus
version when I run it in UNIX mode.

- Paul

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

Re: Make and javac compliler problem in bash

[Jason Tishler]

Paul,

Paul Kinnucan wrote:
> The Windows version of javac does not support paths that BEGIN with a
> forward slash. 

The Windows version of Sun's javac does support paths that begin with
a forward slash.  Although, it will only interpret the path relative to
the current drive letter.

> That is the crux of Jin's complaint.

It certainly will not use Cygwin's mount table -- maybe this is the
crux of Jin's problem.

> Yes, but have you run the Cygnus version of GNU make on Windows/NT?  The
> Cygnus version of GNU make is a version that has been modified to run in
> two modes: Windows mode and  Unix mode. See the Cygnus FAQ for more
> information.  I've generally had the most success running the Cygnus
> version when I run it in UNIX mode.

I have from b17 to b20.1 (and soon on 1.1.2), but only in UNIX mode.

Due to desiring tolerance of \r\n in makefiles, I decided to upgrade
from Cygwin make 3.77 to GNU make 3.79.  It built OOTB without any
fuss.  And since I'm only interested in UNIX mode, I didn't notice any
issues (ie, lack of the Windows mode).

Since Chris was nice enough to release Cygwin make 3.79, I will be
switching back to the make included with the Net distribution as soon
as it is convenient.

Jason

-- 
Jason Tishler
Director, Software Engineering       Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corporation         Fax:   +1 (732) 264-8798
82 Bethany Road, Suite 7             Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA                 WWW:   http://www.dothill.com

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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 04:56 PM 6/12/00 -0400, Jason Tishler wrote:
>Paul,
>
>Paul Kinnucan wrote:
>> The Windows version of javac does not support paths that BEGIN with a
>> forward slash. 
>
>The Windows version of Sun's javac does support paths that begin with
>a forward slash.  Although, it will only interpret the path relative to
>the current drive letter.
>

I didn't know that. I'll check it out. Thanks for pointing this out.

>> That is the crux of Jin's complaint.
>
>It certainly will not use Cygwin's mount table -- maybe this is the
>crux of Jin's problem.
>

Perhaps, although he said nothing to indicate this.

>> Yes, but have you run the Cygnus version of GNU make on Windows/NT?  The
>> Cygnus version of GNU make is a version that has been modified to run in
>> two modes: Windows mode and  Unix mode. See the Cygnus FAQ for more
>> information.  I've generally had the most success running the Cygnus
>> version when I run it in UNIX mode.
>
>I have from b17 to b20.1 (and soon on 1.1.2), but only in UNIX mode.
>
>Due to desiring tolerance of \r\n in makefiles, I decided to upgrade
>from Cygwin make 3.77 to GNU make 3.79.  It built OOTB without any
>fuss.  And since I'm only interested in UNIX mode, I didn't notice any
>issues (ie, lack of the Windows mode).


The Windows mode, which I assume Jin is running, uses the native Windows
command shell to launch executables. I'm guessing that running make in
Windows mode from a Unix shell (i.e., bash) is going to run into standard
I/O problems, which is what Jin seems to be experiencing.

- Paul

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

Re: Make and javac compliler problem in bash

[Woody Jin]

At 02:56 PM 6/12/00 -0400, Paul Kinnucan wrote:
>At 01:26 PM 6/12/00 -0700, Woody Jin wrote:
> >At 11:53 AM 6/12/00 -0400, you wrote:
> >>At 10:21 AM 6/12/00 -0700, you wrote:
> >> ><Problem 1>
> >> >I have a Makefile which compiles java program, if I type
> >> >make, after the screen shows the "javac xxxx.java", and then
> >> >hangs.  The Makefile doesn't need to
> >> >be complex. Any simple Makefile that compiles even HelloWorld.java
> >> >will make the bash console stuck.
>
> >Whatever you do, "make" will stuck with javac.
> >
>
>Did you actually try running make in Unix mode as suggested in the FAQ?


Of course.


> >1) I thought that javac, being a product of Sun Microsystem, (and there 
> exist
> >     its counterpart in Unix) would accept unix style path, which I found
> >is  not
> >     the case. It is better to write an application to accept both styles.
> >Maybe
> >     we should request Sun to do it, letting them know that in Windows there
> >     are many Unix tool users.
> >2) Rather than messing around the backslash and slash all over the places,
> >     and make Makefiles and other shell scripts incompatible with the Unix,
> >     (for example, I want to use the same Makefile on both platform - yes,
> >I have
> >      Sunworkstation on my left side and WindowsNT on my right side),
> >     maybe, bash or some other utility registers applications which requires
> >     windows specific path, and whenever you use unix style path, it
> >automatically
> >     converts to MS-DOS style when the system finally gives the application
> >the path.
> >
>
>Think about what you're asking. If you come up with a general, complete
>solution to the problem of unambiguously converting DOS to Unix paths, and
>vice-versa, without any additional information than what is in the paths
>(don't forget the little problem of drive letters), patent it. A lot of
>people would be very interested in your solution.


It doesn't need to be so genius. For example, try "gvim", which is a
Windows GUI implementation of vim. I can run it using any path names I want
under any environment (whether from MS-DOS console or bash console).

Since cygwin's mount information is in the windows registry,
why do you have any trouble in finding the drive letter ?
If you don't know how, please ask VIM developers.




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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 04:15 PM 6/12/00 -0700, Woody Jin wrote:
>At 02:56 PM 6/12/00 -0400, Paul Kinnucan wrote:
>>At 01:26 PM 6/12/00 -0700, Woody Jin wrote:
>> >At 11:53 AM 6/12/00 -0400, you wrote:
>> >>At 10:21 AM 6/12/00 -0700, you wrote:
>> >> ><Problem 1>
>> >> >I have a Makefile which compiles java program, if I type
>> >> >make, after the screen shows the "javac xxxx.java", and then
>> >> >hangs.  The Makefile doesn't need to
>> >> >be complex. Any simple Makefile that compiles even HelloWorld.java
>> >> >will make the bash console stuck.
>>
>> >Whatever you do, "make" will stuck with javac.
>> >
>>
>>Did you actually try running make in Unix mode as suggested in the FAQ?
>
>
>Of course.
>

And what happened? Were you able to get your makefile to work?

>
>> >1) I thought that javac, being a product of Sun Microsystem, (and there 
>> exist
>> >     its counterpart in Unix) would accept unix style path, which I found
>> >is  not
>> >     the case. It is better to write an application to accept both styles.
>> >Maybe
>> >     we should request Sun to do it, letting them know that in Windows
there
>> >     are many Unix tool users.
>> >2) Rather than messing around the backslash and slash all over the places,
>> >     and make Makefiles and other shell scripts incompatible with the Unix,
>> >     (for example, I want to use the same Makefile on both platform - yes,
>> >I have
>> >      Sunworkstation on my left side and WindowsNT on my right side),
>> >     maybe, bash or some other utility registers applications which
requires
>> >     windows specific path, and whenever you use unix style path, it
>> >automatically
>> >     converts to MS-DOS style when the system finally gives the application
>> >the path.
>> >
>>
>>Think about what you're asking. If you come up with a general, complete
>>solution to the problem of unambiguously converting DOS to Unix paths, and
>>vice-versa, without any additional information than what is in the paths
>>(don't forget the little problem of drive letters), patent it. A lot of
>>people would be very interested in your solution.
>
>
>It doesn't need to be so genius. For example, try "gvim", which is a
>Windows GUI implementation of vim. I can run it using any path names I want
>under any environment (whether from MS-DOS console or bash console).

I said "without any additional information" because you couched your
requirement in a general way. The solution you're suggesting requires that
a user install Cygwin and create a mount table that maps DOS to Unix paths.

I suspect this is a solution of limited appeal to most developers of
Windows applications, including Sun.

It would be great if Sun or someone else ports javac to Cygwin, but at the
moment javac is a native Windows application and that was the situation I
addressed in my original response to you.

- Paul





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

Re: Make and javac compliler problem in bash

[Woody Jin]

At 06:21 PM 6/12/00 -0400, Paul Kinnucan wrote:

> >>Did you actually try running make in Unix mode as suggested in the FAQ?
> >
> >
> >Of course.
> >
>
>And what happened? Were you able to get your makefile to work?


If it had worked, I wouldn't even have bothered to ask here :-)


>I said "without any additional information" because you couched your
>requirement in a general way. The solution you're suggesting requires that
>a user install Cygwin and create a mount table that maps DOS to Unix paths.
>
>I suspect this is a solution of limited appeal to most developers of
>Windows applications, including Sun.
>
>It would be great if Sun or someone else ports javac to Cygwin, but at the
>moment javac is a native Windows application and that was the situation I
>addressed in my original response to you.

Yes, the above is one of the solution that I suggested. Ask them to recognize
the Cygwin community.
But a better way would be to resolve it within Cygwin.
By default, any application that is not in cygwin may be considered as
Windows native application (WNA). So, when you launch WNAs,
cygwin can convert the unix style directory to DOS style.  After the 
conversion,
pass the converted MS-DOS style path to the WNA.
I don't know why you think that this is so impossible or so difficult.
It is just  putting the gvim's functionality into Cygwin.

If an application that is not WNA and not in Cygwin, cygwin may provide
some facility to register those applications. Maybe any application that
uses cygwin.dll may be considered as non-WNAs.





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

Re: Make and javac compliler problem in bash

[J. J. Farrell]

> From: Woody Jin <wjin@houston.geoquest.slb.com>
> 
> But a better way would be to resolve it within Cygwin.
> By default, any application that is not in cygwin may be considered as
> Windows native application (WNA). So, when you launch WNAs,
> cygwin can convert the unix style directory to DOS style.  After the 
> conversion,
> pass the converted MS-DOS style path to the WNA.
> I don't know why you think that this is so impossible or so difficult.
> It is just  putting the gvim's functionality into Cygwin.
> 
> If an application that is not WNA and not in Cygwin, cygwin may provide
> some facility to register those applications. Maybe any application that
> uses cygwin.dll may be considered as non-WNAs.

I'm sure that if you submit patches to fully, robustly, and
unambiguously implement this functionality, they'll be seriously
considered for inclusion.


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

Re: Make and javac compliler problem in bash

[Chris Faylor]

On Mon, Jun 12, 2000 at 04:33:44PM -0700, J. J. Farrell wrote:
>> If an application that is not WNA and not in Cygwin, cygwin may provide
>> some facility to register those applications. Maybe any application that
>> uses cygwin.dll may be considered as non-WNAs.
>
>I'm sure that if you submit patches to fully, robustly, and
>unambiguously implement this functionality, they'll be seriously
>considered for inclusion.

Wow.  I think I'm going to save this response.  I love it!

cgf

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

Re: Make and javac compliler problem in bash

[Carl Thompson]

Woody Jin wrote:

> ...

> But a better way would be to resolve it within Cygwin.  By default,
> any application that is not in cygwin may be considered as Windows
> native application (WNA). So, when you launch WNAs, cygwin can convert
> the unix style directory to DOS style.  After the conversion, pass the
> converted MS-DOS style path to the WNA. I don't know why you think
> that this is so impossible or so difficult.  It is just  putting the
> gvim's functionality into Cygwin.

I like this idea, but I think this functionality would more properly go into
the BASH patches for Cygwin than in the Cygwin DLL itself.  Since BASH (like
all Unix shells) processes metacharacters in command lines itself before
sending them to the application, this would be an natural fit for that
section of the code.  The only real tricky part would be figuring out how to
allow BASH to figure out whether an application is Cygwin or native Windows
(or is that easy?).

> ...

Carl Thompson

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

Re: Make and javac compliler problem in bash

[AJ Reins]

Carl Thompson wrote:
> 
> Woody Jin wrote:
> 
> ...
> 
> I like this idea, but I think this functionality would more properly go into
> the BASH patches for Cygwin than in the Cygwin DLL itself.  Since BASH (like
> all Unix shells) processes metacharacters in command lines itself before
> sending them to the application, this would be an natural fit for that
> section of the code.  The only real tricky part would be figuring out how to
> allow BASH to figure out whether an application is Cygwin or native Windows
> (or is that easy?).
> 

To quote J.J.:
>I'm sure that if you submit patches to fully, robustly, and
>unambiguously implement this functionality, they'll be seriously
>considered for inclusion.

By the way, cygpath is your friend.

-- 
AJ Reins - tbisp<AT>uswest.net -or- tbisp<AT>my-deja.com
Bash maintainer for Cygwin
If there is no time like the present,
 but we never have the time,
does this mean there is no present?

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

Re: Make and javac compliler problem in bash

[Carl Thompson]

AJ Reins wrote:

> ...

> To quote J.J.:
> >I'm sure that if you submit patches to fully, robustly, and
> >unambiguously implement this functionality, they'll be seriously
> >considered for inclusion.

Thanks, but I was capable of reading his message myself.

> ...

> AJ Reins - tbisp<AT>uswest.net -or- tbisp<AT>my-deja.com

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

Re: Make and javac compliler problem in bash

[Woody Jin]

At 06:38 PM 6/12/00 -0700, Carl Thompson wrote:
 >Woody Jin wrote:

 >> But a better way would be to resolve it within Cygwin.  By default,
 >> any application that is not in cygwin may be considered as Windows
 >> native application (WNA). So, when you launch WNAs, cygwin can convert
 >> the unix style directory to DOS style.  After the conversion, pass the
 >> converted MS-DOS style path to the WNA. I don't know why you think
 >> that this is so impossible or so difficult.  It is just  putting the
 >> gvim's functionality into Cygwin.

 >I like this idea, but I think this functionality would more properly go into
 >the BASH patches for Cygwin than in the Cygwin DLL itself.  Since BASH (like
 >all Unix shells) processes metacharacters in command lines itself before
 >sending them to the application, this would be an natural fit for that
 >section of the code.  The only real tricky part would be figuring out how to
 >allow BASH to figure out whether an application is Cygwin or native Windows
 >(or is that easy?).


Carl, this is why I have said from the beginning that some registration
process is required. Obviously, we cannot know.
Even though it is quite rare to use slashes in the arguments
for Windows Native Applications (WNA), we cannot generalize that
every argument that contains slash means directory.
For example, Oracle's SQL Plus will accept the argument like
"user/passwd@instance" - but then who will use this in Windows ?
Everyone I know (includeing myself) clicks on the icon to launch this.

Anyway, one possible way would be like:

% regwna javac       # register javac as wna and convert all arguments
                               #     that contains slashes into MS dos path 
names

% regwna javac -1   # Same as the above but only the first argument
% regwna javac -$   #  only the last argument
% regwna javac -aname "-classpath"   # only the argument for -classpath
% regwna javac - r   #  remove javac as wna.
                               #  etc etc ... I think that the developers 
have a better
                               #    idea of how this may be designed.

I don't think that this will completely solve the problem mathematically,
but I am quite certain that 99.9% of the problems related with path
incomatibility problem will be resolved.

Thanks,

--
Woody Jin


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

Re: Make and javac compliler problem in bash

[Robert Collins]

----- Original Message -----
Subject: Re: Make and javac compliler problem in bash


<....>.  The only real tricky part would be figuring out how to
>  >allow BASH to figure out whether an application is Cygwin or native
Windows
>  >(or is that easy?).
>
>
> Carl, this is why I have said from the beginning that some registration
> process is required.
<...>
> Anyway, one possible way would be like:
>
> % regwna javac       # register javac as wna and convert all arguments
>                                #     that contains slashes into MS dos
path

Couldn't the existing registry keys used to set per cygwin exe flags be used
to accomplish this, with no?/some? minor changes. I believe they are largely
undocumented but I recall Chris saying something...

just my 20c
Rob



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

Re: Make and javac compliler problem in bash

[Chris Faylor]

On Wed, Jun 14, 2000 at 08:55:18AM +1000, Robert Collins wrote:
>Couldn't the existing registry keys used to set per cygwin exe flags be used
>to accomplish this, with no?/some? minor changes. I believe they are largely
>undocumented but I recall Chris saying something...

The flags aren't undocumented.  They are the same as the CYGWIN flags,
parsed by the same mechanism.

cgf

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

Re: Make and javac compliler problem in bash

[Woody Jin]

At 11:42 PM 6/12/00 +0000, Paul Kinnucan wrote:

>That is surprising. Cygnus make works fine in Unix mode for me.
>
>Here is a makefile that I constructed to compile a little test application
>that I've been using to test the Java Development Environment for Emacs:
>
>all:
>         javac -classpath ../../classes -d ../../classes Test.java

In my case, it just doesn't work. It hangs.
I am not sure whether this another problem may be related,
but when I tried "ls /usr/bin", it seemed to hang.
I killed it, and then I brought Windows Task Manager,
and then I executed "ls /usr/bin" again.
WTM's CPU meter went up 100% for 1 min and 20 seconds, after
which "ls /usr/bin" gave me the results. Nothing was running
in background (that is, CPU usage was about 2-4 %).
But then when I tried "make", the CPU meter really didn't go up.




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

Re: Make and javac compliler problem in bash

[Terry Lincoln]

----- Original Message -----
From: "Mortimer, Andy" <andy.mortimer@aeat.co.uk>
To: "Cygwin List" <cygwin@sourceware.cygnus.com>
Sent: Tuesday, June 13, 2000 4:03 AM
Subject: RE: Make and javac compliler problem in bash


> Hi all,
>
> Carl Thompson wrote:
> > Woody Jin wrote:
> >
> > > ...
> >
> > > But a better way would be to resolve it within Cygwin.  By default,
> > > any application that is not in cygwin may be considered as Windows
> > > native application (WNA). So, when you launch WNAs, cygwin
> > can convert
> > > the unix style directory to DOS style.  After the
> > conversion, pass the
> > > converted MS-DOS style path to the WNA. I don't know why you think
> > > that this is so impossible or so difficult.  It is just  putting the
> > > gvim's functionality into Cygwin.
> >
> > I like this idea, but I think this functionality would more
> > properly go into
> > the BASH patches for Cygwin than in the Cygwin DLL itself.
> > [...]
>
> Before taking this much further, please consider the essential difference
between
>   bash$ cmd /c dir
> and
>   bash$ cmd 'c:\' dir
>
> ;-)
That, and any other set of normal DOS (Windows) switches preceded by '/'.
These are
definitely not going to be easily discernible from cygwin-style paths by
anything
less than a full parametric template of all possible switch combinations for
every
DOS command.  This gets worse when things like START are used as they accept
commands
with arguments as there arguments... ad nauseum.

>
> Cheers,
>
> Andy

--- Snip ---
Regards,
W. Terry Lincoln (ICQ# 39362285)
Senior Solution Developer [Linux Registered User #99583]
Ciber corporation http://www.ciber.com
< mailto:WTerryLincoln@engineer.com >
< mailto:lincoln2@rochester.rr.com >
< http://www.geocities.com/terry_lincoln >
< http://www.angelfire.com/ny/TerryLincoln >





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

RE: Make and javac compliler problem in bash

[Mortimer, Andy]

Hi all,

Carl Thompson wrote:
> Woody Jin wrote:
> 
> > ...
> 
> > But a better way would be to resolve it within Cygwin.  By default,
> > any application that is not in cygwin may be considered as Windows
> > native application (WNA). So, when you launch WNAs, cygwin 
> can convert
> > the unix style directory to DOS style.  After the 
> conversion, pass the
> > converted MS-DOS style path to the WNA. I don't know why you think
> > that this is so impossible or so difficult.  It is just  putting the
> > gvim's functionality into Cygwin.
> 
> I like this idea, but I think this functionality would more 
> properly go into
> the BASH patches for Cygwin than in the Cygwin DLL itself.  
> [...]

Before taking this much further, please consider the essential difference between
  bash$ cmd /c dir
and
  bash$ cmd 'c:\' dir

;-)

Cheers,

Andy
***********************************************************************
This transmission contains information which may be confidential and  
which may also be privileged.  It is intended for the named addressee  
only.  Unless you are the named addressee, or authorised to receive it 
on behalf of the addressee you may not copy or use it, or disclose it 
to anyone else.  If you have received this transmission in error please 
contact the sender.  Thank you for your cooperation. 
***********************************************************************

If you would like to find out more about AEA Technology, or how we can
harness the power of science and engineering to help deliver complete
solutions for you, please visit our website at http://www.aeat.co.uk

AEA Technology plc registered office 329 Harwell, Didcot, Oxfordshire OX11 0RA.
Registered in England and Wales, number 3095862.

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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 05:47 PM 6/12/00 -0700, Woody Jin wrote:
>At 06:21 PM 6/12/00 -0400, Paul Kinnucan wrote:
>
>> >>Did you actually try running make in Unix mode as suggested in the FAQ?
>> >
>> >
>> >Of course.
>> >
>>
>>And what happened? Were you able to get your makefile to work?
>
>
>If it had worked, I wouldn't even have bothered to ask here :-)
>

That is surprising. Cygnus make works fine in Unix mode for me.

Here is a makefile that I constructed to compile a little test application
that I've been using to test the Java Development Environment for Emacs:

all:
	javac -classpath ../../classes -d ../../classes Test.java


This makefile compiles Test.java which requires classes stored in the
classes subdirectory of the project directory that contains Test.java. The
makefile is stored in the same directory as Test.java.

This is the output I get when I run Cygnus make in Unix mode on this
makefile from the Cygnus bash shell:

$ make all
javac -classpath ../../classes -d ../../classes Test.java
$

Here is the contents of the classes directory after I run this makefile.

  e:/jmath/classes/jmath:
  total 7
  drwxrwxrwx   2 paulk    123             0 May 12  1998 .
  drwxrwxrwx   2 paulk    123             0 May 12  1998 ..
  -rw-rw-rw-   1 paulk    123           433 Mar 26  1998 JMathError.class
  -rw-rw-rw-   1 paulk    123           484 Dec 11  1999 JMathException.class
  -rw-rw-rw-   1 paulk    123           943 Jun  2 03:12
LinearSystem$InnerClass.class
  -rw-rw-rw-   1 paulk    123          2603 Jun  2 03:12 LinearSystem.class
  -rw-rw-rw-   1 paulk    123           900 Feb  1 01:13 Test$InnerClass.class
  -rw-rw-rw-   1 paulk    123          1542 Jun 13 03:29 Test.class

As you can see, the makefile compiled Test.java as expected.   

- Paul



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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 02:06 PM 6/12/00 -0700, Carl Thompson wrote:
>Paul Kinnucan wrote:
>
>> ...
>
>> >> How could it be otherwise? The Windows version of javac is,
>> >> surprise, a Windows application. Windows applications don't
>> >> recognize Unix paths. Ergo, the Windows version of javac does not
>> >> recognize Unix paths. QED.
>> >
>> >Actually, Win32 _does_ allow path names to use forward slashes as the
>> >directory separator.  The FAQ question you pointed out above even
>> >mentions
>> 
>> But not to BEGIN a path as in Jin's example.
>
>That's certainly not what you wrote above, is it?  Nor is it true.  There is
>nothing inherent in Win32 that says it can't use paths that begin with a "/"
>.  Try pulling up a command prompt and type
>
>   notepad /path/to/a/file

This is simply a shorthand for currentdrive:/path/to/a/file. If you want to
call this a Unix path, fine. I call it a pseudo Unix path of limited
utility. I think it's misleading to generalize from this that some Windows
applications accept Unix paths. In so doing, you're obscuring what's really
going on and implying capabilities that do not exist.

- Paul

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

Re: Make and javac compliler problem in bash

[Carl Thompson]

Paul Kinnucan wrote:

> ...

> >> How could it be otherwise? The Windows version of javac is,
> >> surprise, a Windows application. Windows applications don't
> >> recognize Unix paths. Ergo, the Windows version of javac does not
> >> recognize Unix paths. QED.
> >
> >Actually, Win32 _does_ allow path names to use forward slashes as the
> >directory separator.  The FAQ question you pointed out above even
> >mentions
> 
> But not to BEGIN a path as in Jin's example.

That's certainly not what you wrote above, is it?  Nor is it true.  There is
nothing inherent in Win32 that says it can't use paths that begin with a "/"
.  Try pulling up a command prompt and type

   notepad /path/to/a/file

This works on my system (Windows 2000).  Some programs are written in such a
way that they can handle forward slashes in paths.  "command.com" and
"cmd.exe" have some problems dealing with them, and it looks like "javac"
does too.  This is _not_ a limitation in Windows as you assert, but a
limitation of this programs.

> - Paul

Carl Thompson

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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 11:10 AM 6/12/00 -0700, Carl Thompson wrote:
>Paul Kinnucan wrote:
>> 
>> At 10:21 AM 6/12/00 -0700, you wrote:
>> ><Problem 1>
>> >I have a Makefile which compiles java program, if I type
>> >make, after the screen shows the "javac xxxx.java", and then
>> >hangs.  The Makefile doesn't need to
>> >be complex. Any simple Makefile that compiles even HelloWorld.java
>> >will make the bash console stuck.
>> 
>> See "Why is make behaving badly?" in the Cygnus faq at
>> http://sourceware.cygnus.com/cygwin/faq/
>
>This FAQ entry doesn't seem to address his problem.

According to the FAQ entry, the Cygnus version of make runs in two modes:
Unix and Windows. In Windows mode, make use the Windows command shell  to
launch applications. Thus, in Jin's case, assuming he was running make in
Windows mode (the default), bash launches make which then uses cmd to
launch javac. javac sends it standard out to the Windows command shell and
then make tries forward it to a command shell that it assumes is also cmd
(because it is running in Windows mode) but which is really bash. The
incompatibility between Windows and Unix SIO  seems a reasonable
explanation to me for Jin's problem.

I may be wrong but certainly it wouldn't hurt for Jin to try to run Cygnus
make in Unix mode if he intends to run it under bash.

>
>> ><Problem 2>
>> >It seems to me that javac in Windows will accept directories only in
>> >MS-DOS way. For example, "javac -classpath /a/b/c" won't work.
>> >
>> 
>> How could it be otherwise? The Windows version of javac is, surprise,
>> a Windows application. Windows applications don't recognize Unix
>> paths. Ergo, the Windows version of javac does not recognize Unix
>> paths. QED.
>
>Actually, Win32 _does_ allow path names to use forward slashes as the
>directory separator.  The FAQ question you pointed out above even mentions

But not to BEGIN a path as in Jin's example.


- Paul

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

Re: Make and javac compliler problem in bash

[Carl Thompson]

Paul Kinnucan wrote:
> 
> At 10:21 AM 6/12/00 -0700, you wrote:
> ><Problem 1>
> >I have a Makefile which compiles java program, if I type
> >make, after the screen shows the "javac xxxx.java", and then
> >hangs.  The Makefile doesn't need to
> >be complex. Any simple Makefile that compiles even HelloWorld.java
> >will make the bash console stuck.
> 
> See "Why is make behaving badly?" in the Cygnus faq at
> http://sourceware.cygnus.com/cygwin/faq/

This FAQ entry doesn't seem to address his problem.

> ><Problem 2>
> >It seems to me that javac in Windows will accept directories only in
> >MS-DOS way. For example, "javac -classpath /a/b/c" won't work.
> >
> 
> How could it be otherwise? The Windows version of javac is, surprise,
> a Windows application. Windows applications don't recognize Unix
> paths. Ergo, the Windows version of javac does not recognize Unix
> paths. QED.

Actually, Win32 _does_ allow path names to use forward slashes as the
directory separator.  The FAQ question you pointed out above even mentions
this.  If javac does not allow this, it is not a normal Windows
application.  (Obviously, Win32 does not allow Cygwin style drive
references, but the poster doesn't seem to be talking about that.)

> - Paul

Carl Thompson

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

Re: Make and javac compliler problem in bash

[Paul Kinnucan]

At 10:21 AM 6/12/00 -0700, you wrote:
><Problem 1>
>I have a Makefile which compiles java program, if I type
>make, after the screen shows the "javac xxxx.java", and then
>hangs.  The Makefile doesn't need to
>be complex. Any simple Makefile that compiles even HelloWorld.java
>will make the bash console stuck.

See "Why is make behaving badly?" in the Cygnus faq at
http://sourceware.cygnus.com/cygwin/faq/

><Problem 2>
>It seems to me that javac in Windows will accept directories only in
>MS-DOS way. For example, "javac -classpath /a/b/c" won't work.
>

How could it be otherwise? The Windows version of javac is, surprise, a
Windows application. Windows applications don't recognize Unix paths. Ergo,
the Windows version of javac does not recognize Unix paths. QED.

- Paul 

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

Make and javac compliler problem in bash

[Woody Jin]

<Problem 1>
I have a Makefile which compiles java program, if I type
make, after the screen shows the "javac xxxx.java", and then
hangs.  The Makefile doesn't need to
be complex. Any simple Makefile that compiles even HelloWorld.java
will make the bash console stuck.
<Problem 2>
It seems to me that javac in Windows will accept directories only in
MS-DOS way. For example, "javac -classpath /a/b/c" won't work.

I am running on Windows NT 4.0 with service pack 5.
My Cygwin version: I grabbed the most recent version of *.tar.gz files
of each component from ftp site and installed using the setup program.
I am not sure of the version of this. Maybe the most recent snapshot
of b20 ?
Anyway I am including the  result of  "cygcheck -s -v -r"

Thanks,