Re: Compiled executable differences between 9x and NT

Re: Compiled executable differences between 9x and NT

[N8TM]

In a message dated 3/8/99 6:23:52 AM Pacific Standard Time, jonpryor@vt.edu
writes:

<< What are the *exact* differences between 95 and NT,
 as far as the cygwin environment and compiler are
 concerned? >>

I imagine many of them are dependent on proprietary M$ information.

<<Under 95, it's currently causing
a "blue screen">>

There's one of these when expect crashes in the egcs testsuite on W95.  It
doesn't get that far under NT.

<<Is there anything that is known to break 95 while
working under NT?>>

Of course, but I haven't seen any official list.  One of them is attempting to
build egcs from patch files.  The snapshot releases have made progress on the
vfork failures of the original b20.1 under W95.  There are also things which
work better under W95 than NT.  One of them is catching success/failure
returns from gcc/g++/g77 compiled a.exe.

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

Re: Compiled executable differences between 9x and NT

[N8TM]

In a message dated 3/8/99 6:23:52 AM Pacific Standard Time, jonpryor@vt.edu
writes:

<< What are the *exact* differences between 95 and NT,
 as far as the cygwin environment and compiler are
 concerned? >>

I imagine many of them are dependent on proprietary M$ information.

<<Under 95, it's currently causing
a "blue screen">>

There's one of these when expect crashes in the egcs testsuite on W95.  It
doesn't get that far under NT.

<<Is there anything that is known to break 95 while
working under NT?>>

Of course, but I haven't seen any official list.  One of them is attempting to
build egcs from patch files.  The snapshot releases have made progress on the
vfork failures of the original b20.1 under W95.  There are also things which
work better under W95 than NT.  One of them is catching success/failure
returns from gcc/g++/g77 compiled a.exe.

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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

I suppose I wasn't too clear in my original message.
It was the *application* that was crashing; not Windows.
Granted, it caused a blue-screen in Win95 (which never
gives me a good feeling), but the system was still
running after the fact (Win95 hadn't frozen).
Thus, you're seeing what I was seeing.

So, it looks like I somehow stumbled across an
incompatibility in Win95B that didn't exist before.
Anybody have a later version (OSR2, OSR2.5, Win98, etc.)
to test?  Maybe it's only Win95B that bombs...which would
still be odd, and I'm not sure how it could be solved.
But it would certainly help my curiosity...

As for the download issue...  I hadn't run into that before.
Is "application/zip" close-enough for a tarball?  e.g.

    <a href="gecl.tgz" type="application/zip">source tarball</a>

(I didn't see anything referencing "tar" in the IANA database,
but I saw "application/zip", which seems to fit...)
Hopefully, this should fix the mime type problem -- if you're
accessing from an html page, anyway.  I'm not sure how to set
the mime-type on a file when _directly_ accessing the file,
instead of through an href.  Thoughts?

Thanks,
 - Jon

-----Original Message-----
Glenn Spell <glenn@gs.fay.nc.us> wrote:

>Mumit Khan wrote:
>>
>> Could someone test it on a Win95 B (I know it works on Win95 A) *without*
>> IE 4.0 and see if it still works? I wonder if IE is the deciding factor
>> here.
>
>I run Win95B and I allow *no* traces of any version of MSIE on
>my system. The filesystem and the registry has been wiped clean
>of anything I can find having to do with MSIE.
>
>The 'test.exe' still causes the blue screen but only the program
>terminates. The system does not crash.
>
>BTW, I could not use Netscape or Win95 to download the package
>because the webserver returns a "Content-Type" of "text/plain".
>I grabbed it with lynx on FreeBSD then transferred it to Win95.
>
>-----------------------------------------------------------
> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>
>HTTP/1.1 200 OK
>Date: Fri, 12 Mar 1999 02:48:45 GMT
>Server: Apache/1.2.5
>Last-Modified: Tue, 09 Mar 1999 23:05:34 GMT
>ETag: "1d2e3c-2b688-36e5a93e"
>Content-Length: 177800
>Accept-Ranges: bytes
>Connection: close
>Content-Type: text/plain
>-----------------------------------------------------------
>
>-glenn
>
>--
>            Glenn Spell <glenn@gs.fay.nc.us>
>         Fayetteville, North Carolina, U. S. A.
>  ____________________________________________________
>  ... blue skies ... happy trails ... sweet dreams ...
>
>--
>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: Compiled executable differences between 9x and NT

[Jonathan Pryor]

<inline>

-----Original Message-----
From: Mumit Khan <khan@xraylith.wisc.EDU>
To: Jonathan Pryor <jonpryor>
Cc: Cygwin Mailing List <cygwin@sourceware.cygnus.com>
Date: Tuesday, March 09, 1999 8:08 PM
Subject: Re: Compiled executable differences between 9x and NT 

>"Jonathan Pryor" <jonpryor@vt.edu> writes:
>> A tarball of the source is available at:
>> 
>> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>> 
>> Readme.txt (in the tarball) includes directions on
>> compiling the program, as well as descriptions of 
>> the observed crash under 95.
>
>Jonathan,
>
>I tried out your code (nice work by the way!) on a W95 laptop and I can't
>get it to crash. The specific version listed is 4.00.950a.
>  
>  - cygwin b20.1 + egcs-1.1: Ok
>  - cygwin b20.1 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1 (using Cygwin gcc with -mno-cygwin): Ok.
>
>Perhaps it's the particular version of Win'95 I'm running that makes it
>work? If I remember correctly, I installed SP1 (or whatever it's called)
>on top on the original W95 that this laptop came with.
>
>Regards,
>Mumit

Thank you for the compliment.

It turns out the version of Win95 I'm testing on is 4.00.950 B,
with IE 4.0 SP1.  Furthermore, it had pre-bundled software,
so I have no idea what could be causing a conflict on this
system.

At least I know I'm not going crazy and that it does, in fact,
run under at least one person's copy of Win95, though.

Thanks.  Guess I'll need to clean off the system now...
 - Jon



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

Re: Compiled executable differences between 9x and NT

[Mumit Khan]

Anders Norlander <anorland@hem2.passagen.se> writes:
> 
> I have a w95 system (haven't tested on NT) similar to yours (4.00.950 B
> and IE 4.0 xx) and it crashes for me as well.
> 
> This suggests that there might be some problem with this setup
> or it could be something completely different.
> 

Could someone test it on a Win95 B (I know it works on Win95 A) *without*
IE 4.0 and see if it still works? I wonder if IE is the deciding factor
here.

Regards,
Mumit


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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

I hadn't known about the `.htaccess' file.  Thank you for the
suggestion.  "Correct" headers are now being generated for
the .tgz file (at least according to DJ's HTTP Header Viewer).

Hopefully this should fix any download problems...
 - Jon

Glenn Spell <glenn@gs.fay.nc.us> wrote:


>[This message is off topic but I believe it is of interest.]
>
>Jonathan Pryor wrote:
>> I suppose I wasn't too clear in my original message.
>> It was the *application* that was crashing; not Windows.
>
>I just wanted to make sure you know what I was seeing.
>
>> As for the download issue...  I hadn't run into that before.
>> Is "application/zip" close-enough for a tarball?  e.g.
>> 
>>     <a href="gecl.tgz" type="application/zip">source tarball</a>
>
>Well, I tried again and Netscape 4.05 still corrupts the file on
>download. I assume the problem is still the headers generated
>by your webserver.
>
>> (I didn't see anything referencing "tar" in the IANA database,
>> but I saw "application/zip", which seems to fit...)
>
>I saw that there also. Oddly, either Netscape or WinZip set me
>up in Netscape with "application/x-zip-compressed    .ZIP". I
>changed it to "application/zip" before I tried the download.
>
>> Hopefully, this should fix the mime type problem -- if you're
>> accessing from an html page, anyway.  I'm not sure how to set
>> the mime-type on a file when _directly_ accessing the file,
>> instead of through an href.  Thoughts?
>
>I believe the following method will work regardless of how the
>file is accessed. Since you're using Apache on FreeBSD, just
>make a file named ".htaccess" in that directory with the
>following content:
>
>AddType    application/x-gzip    .gz .tgz
>AddType    application/x-tar     .tar
>
<snip>


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

Compiled executable differences between 9x and NT

[Jonathan Pryor]

(I would try to look this up on the web page, but 
I'm having trouble connecting to cygnus.com at the
moment...)

What are the *exact* differences between 95 and NT,
as far as the cygwin environment and compiler are
concerned?

I would normally assume that if I stick with the
strict ANSI stuff (console I/O), limited to the
extent so that MSVC can also compile it (no great
reliance on the posix api's), that I should be able
to compile an executable on both 95 and NT and have
it behave the same on both.

Unfortunately, I have a program for which this isn't 
happening.  Under NT, it runs as expected in all 
cases.  Under 95, it's currently causing
a "blue screen" with cygwin egcs-1.1.1, and causes
a "This progam has performed an illegal operation
and will be shut down." message under mingw32.

I get these errors if the executable was compiled
and run under 95, or if the executable was compiled
under NT but run under 95.  (The same executable -- 
compiled under either 95 or NT -- runs fine under
NT.)

As a note, it compiles and runs fine with MSVC6.

Any thoughts as to what could be causing the problem?
Is there anything that is known to break 95 while
working under NT?

Thanks,
 - Jon

(Yes, I probably should try to get the whole source
available, but I'm in the process of trying to clean
it up for public distribution, so it may be awhile...)


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

Re: Compiled executable differences between 9x and NT

[Mumit Khan]

"Jonathan Pryor" <jonpryor@vt.edu> writes:
> A tarball of the source is available at:
> 
> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
> 
> Readme.txt (in the tarball) includes directions on
> compiling the program, as well as descriptions of 
> the observed crash under 95.

Jonathan,

I tried out your code (nice work by the way!) on a W95 laptop and I can't
get it to crash. The specific version listed is 4.00.950a.
  
  - cygwin b20.1 + egcs-1.1: Ok
  - cygwin b20.1 + egcs-1.1.1: Ok
  - mingw32 + egcs-1.1.1: Ok
  - mingw32 + egcs-1.1.1 (using Cygwin gcc with -mno-cygwin): Ok.

Perhaps it's the particular version of Win'95 I'm running that makes it
work? If I remember correctly, I installed SP1 (or whatever it's called)
on top on the original W95 that this laptop came with.

Regards,
Mumit


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

Re: Compiled executable differences between 9x and NT

[Suhaib M. Siddiqi]

>As for the download issue...  I hadn't run into that before.
>Is "application/zip" close-enough for a tarball?  e.g.
>
>    <a href="gecl.tgz" type="application/zip">source tarball</a>
>
>(I didn't see anything referencing "tar" in the IANA database,
>but I saw "application/zip", which seems to fit...)
>Hopefully, this should fix the mime type problem -- if you're
>accessing from an html page, anyway.  I'm not sure how to set
>the mime-type on a file when _directly_ accessing the file,
>instead of through an href.  Thoughts?
>


The Netscape latest version (version 4.51) seems to have some bugs which
I noticed.  On Windows Nt and 98 it is corrupting the *.tgz and tar.gz
archives.  Upon extraction or doing a gzunip I get *invalid archive
format* errors.  I discovered this problem with Netscape 4.51 yesterday
when i was downloading *.tar.gz archives from a few different sites.


--SMS


>Thanks,
> - Jon
>
>-----Original Message-----
>Glenn Spell <glenn@gs.fay.nc.us> wrote:
>
>>Mumit Khan wrote:
>>>
>>> Could someone test it on a Win95 B (I know it works on Win95 A)
*without*
>>> IE 4.0 and see if it still works? I wonder if IE is the deciding
factor
>>> here.
>>
>>I run Win95B and I allow *no* traces of any version of MSIE on
>>my system. The filesystem and the registry has been wiped clean
>>of anything I can find having to do with MSIE.
>>
>>The 'test.exe' still causes the blue screen but only the program
>>terminates. The system does not crash.
>>
>>BTW, I could not use Netscape or Win95 to download the package
>>because the webserver returns a "Content-Type" of "text/plain".
>>I grabbed it with lynx on FreeBSD then transferred it to Win95.
>>
>>-----------------------------------------------------------
>> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>>
>>HTTP/1.1 200 OK
>>Date: Fri, 12 Mar 1999 02:48:45 GMT
>>Server: Apache/1.2.5
>>Last-Modified: Tue, 09 Mar 1999 23:05:34 GMT
>>ETag: "1d2e3c-2b688-36e5a93e"
>>Content-Length: 177800
>>Accept-Ranges: bytes
>>Connection: close
>>Content-Type: text/plain
>>-----------------------------------------------------------
>>
>>-glenn
>>
>>--
>>            Glenn Spell <glenn@gs.fay.nc.us>
>>         Fayetteville, North Carolina, U. S. A.
>>  ____________________________________________________
>>  ... blue skies ... happy trails ... sweet dreams ...
>>
>>--
>>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
>
>



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

Re: Compiled executable differences between 9x and NT

[Mumit Khan]

"Jonathan Pryor" <jonpryor@vt.edu> writes:
> 
> I get these errors if the executable was compiled
> and run under 95, or if the executable was compiled
> under NT but run under 95.  (The same executable -- 
> compiled under either 95 or NT -- runs fine under
> NT.)
> 
> As a note, it compiles and runs fine with MSVC6.
> 
> Any thoughts as to what could be causing the problem?
> Is there anything that is known to break 95 while
> working under NT?
> 
> Thanks,
>  - Jon
> 
> (Yes, I probably should try to get the whole source
> available, but I'm in the process of trying to clean
> it up for public distribution, so it may be awhile...)

Without looking at your code, it's almost impossible to tell what is
wrong. Compiler bug? User code bug? Runtime bug? All of these are 
probable, and that's why guesswork is usually a waste of time.

If you can package up the sources, I'll take a look.

Note that W95 is *very* sensitive to memory corruption bugs, and in
many cases, these codes will run seemingly fine on Linux, NT etc.

Regards,
Mumit


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

Re: Compiled executable differences between 9x and NT

[Geoffrey Noer]

On Mon, Mar 08, 1999, Jonathan Pryor wrote:
>
> What are the *exact* differences between 95 and NT,
> as far as the cygwin environment and compiler are
> concerned?
> 
> I would normally assume that if I stick with the
> strict ANSI stuff (console I/O), limited to the
> extent so that MSVC can also compile it (no great
> reliance on the posix api's), that I should be able
> to compile an executable on both 95 and NT and have
> it behave the same on both.

9x and NT have different sets of bugs and features.  Cygwin
checks which OS is running and uses this info to provide the
a Unix layer that should make Cygwin applications run
the same under either OS.  (Of course while this is the
goal, sometimes this doesn't quite happen).

> Unfortunately, I have a program for which this isn't 
> happening.  Under NT, it runs as expected in all 
> cases.  Under 95, it's currently causing
> a "blue screen" with cygwin egcs-1.1.1, and causes
> a "This progam has performed an illegal operation
> and will be shut down." message under mingw32.
> 
> I get these errors if the executable was compiled
> and run under 95, or if the executable was compiled
> under NT but run under 95.  (The same executable -- 
> compiled under either 95 or NT -- runs fine under
> NT.)
> 
> As a note, it compiles and runs fine with MSVC6.

Ah, you're using mingw.  So Cygwin is probably
not relevant.  Hmmm, well, I've noticed that Windows
9x often is more sensitive to buggy exes.  Perhaps
you're running into a compiler issue?

-- 
Geoffrey Noer		Email: noer@cygnus.com
Cygnus Solutions

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

Re: Compiled executable differences between 9x and NT

[Glenn Spell]

Mumit Khan wrote:
> 
> Could someone test it on a Win95 B (I know it works on Win95 A) *without*
> IE 4.0 and see if it still works? I wonder if IE is the deciding factor
> here.

I run Win95B and I allow *no* traces of any version of MSIE on
my system. The filesystem and the registry has been wiped clean
of anything I can find having to do with MSIE.

The 'test.exe' still causes the blue screen but only the program
terminates. The system does not crash.

BTW, I could not use Netscape or Win95 to download the package
because the webserver returns a "Content-Type" of "text/plain".
I grabbed it with lynx on FreeBSD then transferred it to Win95.

-----------------------------------------------------------
http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz

HTTP/1.1 200 OK
Date: Fri, 12 Mar 1999 02:48:45 GMT
Server: Apache/1.2.5
Last-Modified: Tue, 09 Mar 1999 23:05:34 GMT
ETag: "1d2e3c-2b688-36e5a93e"
Content-Length: 177800
Accept-Ranges: bytes
Connection: close
Content-Type: text/plain
-----------------------------------------------------------

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...

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

Re: Compiled executable differences between 9x and NT

[neud]

Mr. Kuehn never works by us. Please actualize your Mail-Listings.

Thank you






Mumit Khan schrieb:

> Anders Norlander <anorland@hem2.passagen.se> writes:
> >
> > I have a w95 system (haven't tested on NT) similar to yours (4.00.950 B
> > and IE 4.0 xx) and it crashes for me as well.
> >
> > This suggests that there might be some problem with this setup
> > or it could be something completely different.
> >
>
> Could someone test it on a Win95 B (I know it works on Win95 A) *without*
> IE 4.0 and see if it still works? I wonder if IE is the deciding factor
> here.
>
> Regards,
> Mumit
>
> --
> 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: Compiled executable differences between 9x and NT

[Jonathan Pryor]

Does your W95 laptop have IE installed?  Given the 
results from Anders Norlander, it's possible that
IE is causing problems...

Thanks,
 - Jon

-----Original Message-----
From: Mumit Khan <khan@xraylith.wisc.EDU>
To: Jonathan Pryor <jonpryor>
Cc: Cygwin Mailing List <cygwin@sourceware.cygnus.com>
Date: Tuesday, March 09, 1999 8:08 PM
Subject: Re: Compiled executable differences between 9x and NT 


>"Jonathan Pryor" <jonpryor@vt.edu> writes:
>> A tarball of the source is available at:
>> 
>> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>> 
>> Readme.txt (in the tarball) includes directions on
>> compiling the program, as well as descriptions of 
>> the observed crash under 95.
>
>Jonathan,
>
>I tried out your code (nice work by the way!) on a W95 laptop and I can't
>get it to crash. The specific version listed is 4.00.950a.
>  
>  - cygwin b20.1 + egcs-1.1: Ok
>  - cygwin b20.1 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1 (using Cygwin gcc with -mno-cygwin): Ok.
>
>Perhaps it's the particular version of Win'95 I'm running that makes it
>work? If I remember correctly, I installed SP1 (or whatever it's called)
>on top on the original W95 that this laptop came with.
>
>Regards,
>Mumit



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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

A tarball of the source is available at:

http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz

Readme.txt (in the tarball) includes directions on
compiling the program, as well as descriptions of 
the observed crash under 95.

<snip>

>9x and NT have different sets of bugs and features.  Cygwin
>checks which OS is running and uses this info to provide the
>a Unix layer that should make Cygwin applications run
>the same under either OS.  (Of course while this is the
>goal, sometimes this doesn't quite happen).


Cygwin isn't there yet, but it's getting close.  Thank you
for the effort.

<snip>

>Ah, you're using mingw.  So Cygwin is probably
>not relevant.  Hmmm, well, I've noticed that Windows
>9x often is more sensitive to buggy exes.  Perhaps
>you're running into a compiler issue?


Actually, I'm trying to use as many compilers as I can.
Thus, I'm compiling with cygwin, mingw32, and MSVC.
(I would also try DJGPP, but I had some configuration
issues I still need to deal with...)

As for compiler issues...  I seem to have run into a fair
number of them trying to develop the program.  I sent out
a prior message detailing 3 template-related bugs to the
cygwin list (as well as to egcs-bugs).  I believe I've
worked around them, but it's possible I missed a few.

>-- 
>Geoffrey Noer Email: noer@cygnus.com
>Cygnus Solutions
>


Thanks,
 - Jon



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

Re: Compiled executable differences between 9x and NT

[Anders Norlander]

Jonathan Pryor wrote:
> It turns out the version of Win95 I'm testing on is 4.00.950 B,
> with IE 4.0 SP1.  Furthermore, it had pre-bundled software,
> so I have no idea what could be causing a conflict on this
> system.
> 
> At least I know I'm not going crazy and that it does, in fact,
> run under at least one person's copy of Win95, though.

I have a w95 system (haven't tested on NT) similar to yours (4.00.950 B
and IE 4.0 xx) and it crashes for me as well.

This suggests that there might be some problem with this setup
or it could be something completely different.

Anders

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

Re: Compiled executable differences between 9x and NT

[Glenn Spell]

[This message is off topic but I believe it is of interest.]

Jonathan Pryor wrote:
> I suppose I wasn't too clear in my original message.
> It was the *application* that was crashing; not Windows.

I just wanted to make sure you know what I was seeing.

> As for the download issue...  I hadn't run into that before.
> Is "application/zip" close-enough for a tarball?  e.g.
> 
>     <a href="gecl.tgz" type="application/zip">source tarball</a>

Well, I tried again and Netscape 4.05 still corrupts the file on
download. I assume the problem is still the headers generated
by your webserver.

> (I didn't see anything referencing "tar" in the IANA database,
> but I saw "application/zip", which seems to fit...)

I saw that there also. Oddly, either Netscape or WinZip set me
up in Netscape with "application/x-zip-compressed    .ZIP". I
changed it to "application/zip" before I tried the download.

> Hopefully, this should fix the mime type problem -- if you're
> accessing from an html page, anyway.  I'm not sure how to set
> the mime-type on a file when _directly_ accessing the file,
> instead of through an href.  Thoughts?

I believe the following method will work regardless of how the
file is accessed. Since you're using Apache on FreeBSD, just
make a file named ".htaccess" in that directory with the
following content:

AddType    application/x-gzip    .gz .tgz
AddType    application/x-tar     .tar

File permissions must be correct and there may need to be
other directives in the file depending on how Apache is
configured. Check with the Head System Admin (Daniel Hagan)
there or the WWW Group <www@acm.vt.edu> if you need help.

Actually, those types are so common they should have been
set up in the global config files by the admin. (I searched
for examples of ".tgz" on the Net but didn't find any. I
don't think what follows "application/" is critical... only
that the server recognizes the extension ".tgz" and prints
a Content-Type header that results in a binary download.)

To check the headers sent by your server you can use DJ's
"HTTP Header Viewer" at:
    < http://www.delorie.com/web/headers.html >

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...

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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

<inline>
-----Original Message-----
From: N8TM@aol.com <N8TM@aol.com>
To: jonpryor <jonpryor>; cygwin@sourceware.cygnus.com
<cygwin@sourceware.cygnus.com>
Date: Monday, March 08, 1999 9:37 PM
Subject: Re: Compiled executable differences between 9x and NT


>In a message dated 3/8/99 6:23:52 AM Pacific Standard Time, jonpryor@vt.edu
>writes:
>
><< What are the *exact* differences between 95 and NT,
> as far as the cygwin environment and compiler are
> concerned? >>
>
>I imagine many of them are dependent on proprietary M$ information.

Differences between 95 and NT would be proprietary information.
I fail to see why differences in the behavior of cygwin-compiled
executables would be Microsoft proprietary, though...  At the very
least, I would expect someone to have an idea of what (programs,
operations, commands, source code, etc.), in general, tends to
"break" 95 while working fine under NT.

><<Under 95, it's currently causing
>a "blue screen">>
>
>There's one of these when expect crashes in the egcs testsuite on W95.  It
>doesn't get that far under NT.

What causes the crash in the egcs testsuite?  Which test?  Why does
it break (if known)?

><<Is there anything that is known to break 95 while
>working under NT?>>
>
>Of course, but I haven't seen any official list.  One of them is attempting
to
>build egcs from patch files.  The snapshot releases have made progress on
the
>vfork failures of the original b20.1 under W95.  There are also things
which
>work better under W95 than NT.  One of them is catching success/failure
>returns from gcc/g++/g77 compiled a.exe.

I suppose a better question would be: What C/C++ source has a
tendancy of segfaulting under 95, but working fine under NT?

I'd like to narrow down whether this is a problem with the
runtime under 95, or a problem with the OS itself.  Either way,
if I know what source is "unsafe" under 95, I can try to re-write
my code to work safely under 95 as well as NT.  But until I know
what issues to look out for, re-writing isn't an issue.

Thanks,
- Jon


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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

I hadn't known about the `.htaccess' file.  Thank you for the
suggestion.  "Correct" headers are now being generated for
the .tgz file (at least according to DJ's HTTP Header Viewer).

Hopefully this should fix any download problems...
 - Jon

Glenn Spell <glenn@gs.fay.nc.us> wrote:


>[This message is off topic but I believe it is of interest.]
>
>Jonathan Pryor wrote:
>> I suppose I wasn't too clear in my original message.
>> It was the *application* that was crashing; not Windows.
>
>I just wanted to make sure you know what I was seeing.
>
>> As for the download issue...  I hadn't run into that before.
>> Is "application/zip" close-enough for a tarball?  e.g.
>> 
>>     <a href="gecl.tgz" type="application/zip">source tarball</a>
>
>Well, I tried again and Netscape 4.05 still corrupts the file on
>download. I assume the problem is still the headers generated
>by your webserver.
>
>> (I didn't see anything referencing "tar" in the IANA database,
>> but I saw "application/zip", which seems to fit...)
>
>I saw that there also. Oddly, either Netscape or WinZip set me
>up in Netscape with "application/x-zip-compressed    .ZIP". I
>changed it to "application/zip" before I tried the download.
>
>> Hopefully, this should fix the mime type problem -- if you're
>> accessing from an html page, anyway.  I'm not sure how to set
>> the mime-type on a file when _directly_ accessing the file,
>> instead of through an href.  Thoughts?
>
>I believe the following method will work regardless of how the
>file is accessed. Since you're using Apache on FreeBSD, just
>make a file named ".htaccess" in that directory with the
>following content:
>
>AddType    application/x-gzip    .gz .tgz
>AddType    application/x-tar     .tar
>
<snip>


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

Re: Compiled executable differences between 9x and NT

[Glenn Spell]

[This message is off topic but I believe it is of interest.]

Jonathan Pryor wrote:
> I suppose I wasn't too clear in my original message.
> It was the *application* that was crashing; not Windows.

I just wanted to make sure you know what I was seeing.

> As for the download issue...  I hadn't run into that before.
> Is "application/zip" close-enough for a tarball?  e.g.
> 
>     <a href="gecl.tgz" type="application/zip">source tarball</a>

Well, I tried again and Netscape 4.05 still corrupts the file on
download. I assume the problem is still the headers generated
by your webserver.

> (I didn't see anything referencing "tar" in the IANA database,
> but I saw "application/zip", which seems to fit...)

I saw that there also. Oddly, either Netscape or WinZip set me
up in Netscape with "application/x-zip-compressed    .ZIP". I
changed it to "application/zip" before I tried the download.

> Hopefully, this should fix the mime type problem -- if you're
> accessing from an html page, anyway.  I'm not sure how to set
> the mime-type on a file when _directly_ accessing the file,
> instead of through an href.  Thoughts?

I believe the following method will work regardless of how the
file is accessed. Since you're using Apache on FreeBSD, just
make a file named ".htaccess" in that directory with the
following content:

AddType    application/x-gzip    .gz .tgz
AddType    application/x-tar     .tar

File permissions must be correct and there may need to be
other directives in the file depending on how Apache is
configured. Check with the Head System Admin (Daniel Hagan)
there or the WWW Group <www@acm.vt.edu> if you need help.

Actually, those types are so common they should have been
set up in the global config files by the admin. (I searched
for examples of ".tgz" on the Net but didn't find any. I
don't think what follows "application/" is critical... only
that the server recognizes the extension ".tgz" and prints
a Content-Type header that results in a binary download.)

To check the headers sent by your server you can use DJ's
"HTTP Header Viewer" at:
    < http://www.delorie.com/web/headers.html >

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...

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

Re: Compiled executable differences between 9x and NT

[Suhaib M. Siddiqi]

>As for the download issue...  I hadn't run into that before.
>Is "application/zip" close-enough for a tarball?  e.g.
>
>    <a href="gecl.tgz" type="application/zip">source tarball</a>
>
>(I didn't see anything referencing "tar" in the IANA database,
>but I saw "application/zip", which seems to fit...)
>Hopefully, this should fix the mime type problem -- if you're
>accessing from an html page, anyway.  I'm not sure how to set
>the mime-type on a file when _directly_ accessing the file,
>instead of through an href.  Thoughts?
>


The Netscape latest version (version 4.51) seems to have some bugs which
I noticed.  On Windows Nt and 98 it is corrupting the *.tgz and tar.gz
archives.  Upon extraction or doing a gzunip I get *invalid archive
format* errors.  I discovered this problem with Netscape 4.51 yesterday
when i was downloading *.tar.gz archives from a few different sites.


--SMS


>Thanks,
> - Jon
>
>-----Original Message-----
>Glenn Spell <glenn@gs.fay.nc.us> wrote:
>
>>Mumit Khan wrote:
>>>
>>> Could someone test it on a Win95 B (I know it works on Win95 A)
*without*
>>> IE 4.0 and see if it still works? I wonder if IE is the deciding
factor
>>> here.
>>
>>I run Win95B and I allow *no* traces of any version of MSIE on
>>my system. The filesystem and the registry has been wiped clean
>>of anything I can find having to do with MSIE.
>>
>>The 'test.exe' still causes the blue screen but only the program
>>terminates. The system does not crash.
>>
>>BTW, I could not use Netscape or Win95 to download the package
>>because the webserver returns a "Content-Type" of "text/plain".
>>I grabbed it with lynx on FreeBSD then transferred it to Win95.
>>
>>-----------------------------------------------------------
>> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>>
>>HTTP/1.1 200 OK
>>Date: Fri, 12 Mar 1999 02:48:45 GMT
>>Server: Apache/1.2.5
>>Last-Modified: Tue, 09 Mar 1999 23:05:34 GMT
>>ETag: "1d2e3c-2b688-36e5a93e"
>>Content-Length: 177800
>>Accept-Ranges: bytes
>>Connection: close
>>Content-Type: text/plain
>>-----------------------------------------------------------
>>
>>-glenn
>>
>>--
>>            Glenn Spell <glenn@gs.fay.nc.us>
>>         Fayetteville, North Carolina, U. S. A.
>>  ____________________________________________________
>>  ... blue skies ... happy trails ... sweet dreams ...
>>
>>--
>>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
>
>



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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

I suppose I wasn't too clear in my original message.
It was the *application* that was crashing; not Windows.
Granted, it caused a blue-screen in Win95 (which never
gives me a good feeling), but the system was still
running after the fact (Win95 hadn't frozen).
Thus, you're seeing what I was seeing.

So, it looks like I somehow stumbled across an
incompatibility in Win95B that didn't exist before.
Anybody have a later version (OSR2, OSR2.5, Win98, etc.)
to test?  Maybe it's only Win95B that bombs...which would
still be odd, and I'm not sure how it could be solved.
But it would certainly help my curiosity...

As for the download issue...  I hadn't run into that before.
Is "application/zip" close-enough for a tarball?  e.g.

    <a href="gecl.tgz" type="application/zip">source tarball</a>

(I didn't see anything referencing "tar" in the IANA database,
but I saw "application/zip", which seems to fit...)
Hopefully, this should fix the mime type problem -- if you're
accessing from an html page, anyway.  I'm not sure how to set
the mime-type on a file when _directly_ accessing the file,
instead of through an href.  Thoughts?

Thanks,
 - Jon

-----Original Message-----
Glenn Spell <glenn@gs.fay.nc.us> wrote:

>Mumit Khan wrote:
>>
>> Could someone test it on a Win95 B (I know it works on Win95 A) *without*
>> IE 4.0 and see if it still works? I wonder if IE is the deciding factor
>> here.
>
>I run Win95B and I allow *no* traces of any version of MSIE on
>my system. The filesystem and the registry has been wiped clean
>of anything I can find having to do with MSIE.
>
>The 'test.exe' still causes the blue screen but only the program
>terminates. The system does not crash.
>
>BTW, I could not use Netscape or Win95 to download the package
>because the webserver returns a "Content-Type" of "text/plain".
>I grabbed it with lynx on FreeBSD then transferred it to Win95.
>
>-----------------------------------------------------------
> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>
>HTTP/1.1 200 OK
>Date: Fri, 12 Mar 1999 02:48:45 GMT
>Server: Apache/1.2.5
>Last-Modified: Tue, 09 Mar 1999 23:05:34 GMT
>ETag: "1d2e3c-2b688-36e5a93e"
>Content-Length: 177800
>Accept-Ranges: bytes
>Connection: close
>Content-Type: text/plain
>-----------------------------------------------------------
>
>-glenn
>
>--
>            Glenn Spell <glenn@gs.fay.nc.us>
>         Fayetteville, North Carolina, U. S. A.
>  ____________________________________________________
>  ... blue skies ... happy trails ... sweet dreams ...
>
>--
>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: Compiled executable differences between 9x and NT

[Glenn Spell]

Mumit Khan wrote:
> 
> Could someone test it on a Win95 B (I know it works on Win95 A) *without*
> IE 4.0 and see if it still works? I wonder if IE is the deciding factor
> here.

I run Win95B and I allow *no* traces of any version of MSIE on
my system. The filesystem and the registry has been wiped clean
of anything I can find having to do with MSIE.

The 'test.exe' still causes the blue screen but only the program
terminates. The system does not crash.

BTW, I could not use Netscape or Win95 to download the package
because the webserver returns a "Content-Type" of "text/plain".
I grabbed it with lynx on FreeBSD then transferred it to Win95.

-----------------------------------------------------------
http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz

HTTP/1.1 200 OK
Date: Fri, 12 Mar 1999 02:48:45 GMT
Server: Apache/1.2.5
Last-Modified: Tue, 09 Mar 1999 23:05:34 GMT
ETag: "1d2e3c-2b688-36e5a93e"
Content-Length: 177800
Accept-Ranges: bytes
Connection: close
Content-Type: text/plain
-----------------------------------------------------------

-glenn

-- 
            Glenn Spell <glenn@gs.fay.nc.us>
         Fayetteville, North Carolina, U. S. A.
  ____________________________________________________
  ... blue skies ... happy trails ... sweet dreams ...

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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

Does your W95 laptop have IE installed?  Given the 
results from Anders Norlander, it's possible that
IE is causing problems...

Thanks,
 - Jon

-----Original Message-----
From: Mumit Khan <khan@xraylith.wisc.EDU>
To: Jonathan Pryor <jonpryor>
Cc: Cygwin Mailing List <cygwin@sourceware.cygnus.com>
Date: Tuesday, March 09, 1999 8:08 PM
Subject: Re: Compiled executable differences between 9x and NT 


>"Jonathan Pryor" <jonpryor@vt.edu> writes:
>> A tarball of the source is available at:
>> 
>> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>> 
>> Readme.txt (in the tarball) includes directions on
>> compiling the program, as well as descriptions of 
>> the observed crash under 95.
>
>Jonathan,
>
>I tried out your code (nice work by the way!) on a W95 laptop and I can't
>get it to crash. The specific version listed is 4.00.950a.
>  
>  - cygwin b20.1 + egcs-1.1: Ok
>  - cygwin b20.1 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1 (using Cygwin gcc with -mno-cygwin): Ok.
>
>Perhaps it's the particular version of Win'95 I'm running that makes it
>work? If I remember correctly, I installed SP1 (or whatever it's called)
>on top on the original W95 that this laptop came with.
>
>Regards,
>Mumit



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

Re: Compiled executable differences between 9x and NT

[neud]

Mr. Kuehn never works by us. Please actualize your Mail-Listings.

Thank you






Mumit Khan schrieb:

> Anders Norlander <anorland@hem2.passagen.se> writes:
> >
> > I have a w95 system (haven't tested on NT) similar to yours (4.00.950 B
> > and IE 4.0 xx) and it crashes for me as well.
> >
> > This suggests that there might be some problem with this setup
> > or it could be something completely different.
> >
>
> Could someone test it on a Win95 B (I know it works on Win95 A) *without*
> IE 4.0 and see if it still works? I wonder if IE is the deciding factor
> here.
>
> Regards,
> Mumit
>
> --
> 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: Compiled executable differences between 9x and NT

[Mumit Khan]

Anders Norlander <anorland@hem2.passagen.se> writes:
> 
> I have a w95 system (haven't tested on NT) similar to yours (4.00.950 B
> and IE 4.0 xx) and it crashes for me as well.
> 
> This suggests that there might be some problem with this setup
> or it could be something completely different.
> 

Could someone test it on a Win95 B (I know it works on Win95 A) *without*
IE 4.0 and see if it still works? I wonder if IE is the deciding factor
here.

Regards,
Mumit


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

Re: Compiled executable differences between 9x and NT

[Anders Norlander]

Jonathan Pryor wrote:
> It turns out the version of Win95 I'm testing on is 4.00.950 B,
> with IE 4.0 SP1.  Furthermore, it had pre-bundled software,
> so I have no idea what could be causing a conflict on this
> system.
> 
> At least I know I'm not going crazy and that it does, in fact,
> run under at least one person's copy of Win95, though.

I have a w95 system (haven't tested on NT) similar to yours (4.00.950 B
and IE 4.0 xx) and it crashes for me as well.

This suggests that there might be some problem with this setup
or it could be something completely different.

Anders

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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

<inline>

-----Original Message-----
From: Mumit Khan <khan@xraylith.wisc.EDU>
To: Jonathan Pryor <jonpryor>
Cc: Cygwin Mailing List <cygwin@sourceware.cygnus.com>
Date: Tuesday, March 09, 1999 8:08 PM
Subject: Re: Compiled executable differences between 9x and NT 

>"Jonathan Pryor" <jonpryor@vt.edu> writes:
>> A tarball of the source is available at:
>> 
>> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
>> 
>> Readme.txt (in the tarball) includes directions on
>> compiling the program, as well as descriptions of 
>> the observed crash under 95.
>
>Jonathan,
>
>I tried out your code (nice work by the way!) on a W95 laptop and I can't
>get it to crash. The specific version listed is 4.00.950a.
>  
>  - cygwin b20.1 + egcs-1.1: Ok
>  - cygwin b20.1 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1: Ok
>  - mingw32 + egcs-1.1.1 (using Cygwin gcc with -mno-cygwin): Ok.
>
>Perhaps it's the particular version of Win'95 I'm running that makes it
>work? If I remember correctly, I installed SP1 (or whatever it's called)
>on top on the original W95 that this laptop came with.
>
>Regards,
>Mumit

Thank you for the compliment.

It turns out the version of Win95 I'm testing on is 4.00.950 B,
with IE 4.0 SP1.  Furthermore, it had pre-bundled software,
so I have no idea what could be causing a conflict on this
system.

At least I know I'm not going crazy and that it does, in fact,
run under at least one person's copy of Win95, though.

Thanks.  Guess I'll need to clean off the system now...
 - Jon



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

Re: Compiled executable differences between 9x and NT

[Mumit Khan]

"Jonathan Pryor" <jonpryor@vt.edu> writes:
> A tarball of the source is available at:
> 
> http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz
> 
> Readme.txt (in the tarball) includes directions on
> compiling the program, as well as descriptions of 
> the observed crash under 95.

Jonathan,

I tried out your code (nice work by the way!) on a W95 laptop and I can't
get it to crash. The specific version listed is 4.00.950a.
  
  - cygwin b20.1 + egcs-1.1: Ok
  - cygwin b20.1 + egcs-1.1.1: Ok
  - mingw32 + egcs-1.1.1: Ok
  - mingw32 + egcs-1.1.1 (using Cygwin gcc with -mno-cygwin): Ok.

Perhaps it's the particular version of Win'95 I'm running that makes it
work? If I remember correctly, I installed SP1 (or whatever it's called)
on top on the original W95 that this laptop came with.

Regards,
Mumit


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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

A tarball of the source is available at:

http://www.acm.vt.edu/~jonpryor/Projects/gecl.tgz

Readme.txt (in the tarball) includes directions on
compiling the program, as well as descriptions of 
the observed crash under 95.

<snip>

>9x and NT have different sets of bugs and features.  Cygwin
>checks which OS is running and uses this info to provide the
>a Unix layer that should make Cygwin applications run
>the same under either OS.  (Of course while this is the
>goal, sometimes this doesn't quite happen).


Cygwin isn't there yet, but it's getting close.  Thank you
for the effort.

<snip>

>Ah, you're using mingw.  So Cygwin is probably
>not relevant.  Hmmm, well, I've noticed that Windows
>9x often is more sensitive to buggy exes.  Perhaps
>you're running into a compiler issue?


Actually, I'm trying to use as many compilers as I can.
Thus, I'm compiling with cygwin, mingw32, and MSVC.
(I would also try DJGPP, but I had some configuration
issues I still need to deal with...)

As for compiler issues...  I seem to have run into a fair
number of them trying to develop the program.  I sent out
a prior message detailing 3 template-related bugs to the
cygwin list (as well as to egcs-bugs).  I believe I've
worked around them, but it's possible I missed a few.

>-- 
>Geoffrey Noer Email: noer@cygnus.com
>Cygnus Solutions
>


Thanks,
 - Jon



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

Re: Compiled executable differences between 9x and NT

[Mumit Khan]

"Jonathan Pryor" <jonpryor@vt.edu> writes:
> 
> I get these errors if the executable was compiled
> and run under 95, or if the executable was compiled
> under NT but run under 95.  (The same executable -- 
> compiled under either 95 or NT -- runs fine under
> NT.)
> 
> As a note, it compiles and runs fine with MSVC6.
> 
> Any thoughts as to what could be causing the problem?
> Is there anything that is known to break 95 while
> working under NT?
> 
> Thanks,
>  - Jon
> 
> (Yes, I probably should try to get the whole source
> available, but I'm in the process of trying to clean
> it up for public distribution, so it may be awhile...)

Without looking at your code, it's almost impossible to tell what is
wrong. Compiler bug? User code bug? Runtime bug? All of these are 
probable, and that's why guesswork is usually a waste of time.

If you can package up the sources, I'll take a look.

Note that W95 is *very* sensitive to memory corruption bugs, and in
many cases, these codes will run seemingly fine on Linux, NT etc.

Regards,
Mumit


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

Re: Compiled executable differences between 9x and NT

[Geoffrey Noer]

On Mon, Mar 08, 1999, Jonathan Pryor wrote:
>
> What are the *exact* differences between 95 and NT,
> as far as the cygwin environment and compiler are
> concerned?
> 
> I would normally assume that if I stick with the
> strict ANSI stuff (console I/O), limited to the
> extent so that MSVC can also compile it (no great
> reliance on the posix api's), that I should be able
> to compile an executable on both 95 and NT and have
> it behave the same on both.

9x and NT have different sets of bugs and features.  Cygwin
checks which OS is running and uses this info to provide the
a Unix layer that should make Cygwin applications run
the same under either OS.  (Of course while this is the
goal, sometimes this doesn't quite happen).

> Unfortunately, I have a program for which this isn't 
> happening.  Under NT, it runs as expected in all 
> cases.  Under 95, it's currently causing
> a "blue screen" with cygwin egcs-1.1.1, and causes
> a "This progam has performed an illegal operation
> and will be shut down." message under mingw32.
> 
> I get these errors if the executable was compiled
> and run under 95, or if the executable was compiled
> under NT but run under 95.  (The same executable -- 
> compiled under either 95 or NT -- runs fine under
> NT.)
> 
> As a note, it compiles and runs fine with MSVC6.

Ah, you're using mingw.  So Cygwin is probably
not relevant.  Hmmm, well, I've noticed that Windows
9x often is more sensitive to buggy exes.  Perhaps
you're running into a compiler issue?

-- 
Geoffrey Noer		Email: noer@cygnus.com
Cygnus Solutions

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

Re: Compiled executable differences between 9x and NT

[Jonathan Pryor]

<inline>
-----Original Message-----
From: N8TM@aol.com <N8TM@aol.com>
To: jonpryor <jonpryor>; cygwin@sourceware.cygnus.com
<cygwin@sourceware.cygnus.com>
Date: Monday, March 08, 1999 9:37 PM
Subject: Re: Compiled executable differences between 9x and NT


>In a message dated 3/8/99 6:23:52 AM Pacific Standard Time, jonpryor@vt.edu
>writes:
>
><< What are the *exact* differences between 95 and NT,
> as far as the cygwin environment and compiler are
> concerned? >>
>
>I imagine many of them are dependent on proprietary M$ information.

Differences between 95 and NT would be proprietary information.
I fail to see why differences in the behavior of cygwin-compiled
executables would be Microsoft proprietary, though...  At the very
least, I would expect someone to have an idea of what (programs,
operations, commands, source code, etc.), in general, tends to
"break" 95 while working fine under NT.

><<Under 95, it's currently causing
>a "blue screen">>
>
>There's one of these when expect crashes in the egcs testsuite on W95.  It
>doesn't get that far under NT.

What causes the crash in the egcs testsuite?  Which test?  Why does
it break (if known)?

><<Is there anything that is known to break 95 while
>working under NT?>>
>
>Of course, but I haven't seen any official list.  One of them is attempting
to
>build egcs from patch files.  The snapshot releases have made progress on
the
>vfork failures of the original b20.1 under W95.  There are also things
which
>work better under W95 than NT.  One of them is catching success/failure
>returns from gcc/g++/g77 compiled a.exe.

I suppose a better question would be: What C/C++ source has a
tendancy of segfaulting under 95, but working fine under NT?

I'd like to narrow down whether this is a problem with the
runtime under 95, or a problem with the OS itself.  Either way,
if I know what source is "unsafe" under 95, I can try to re-write
my code to work safely under 95 as well as NT.  But until I know
what issues to look out for, re-writing isn't an issue.

Thanks,
- Jon


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

Compiled executable differences between 9x and NT

[Jonathan Pryor]

(I would try to look this up on the web page, but 
I'm having trouble connecting to cygnus.com at the
moment...)

What are the *exact* differences between 95 and NT,
as far as the cygwin environment and compiler are
concerned?

I would normally assume that if I stick with the
strict ANSI stuff (console I/O), limited to the
extent so that MSVC can also compile it (no great
reliance on the posix api's), that I should be able
to compile an executable on both 95 and NT and have
it behave the same on both.

Unfortunately, I have a program for which this isn't 
happening.  Under NT, it runs as expected in all 
cases.  Under 95, it's currently causing
a "blue screen" with cygwin egcs-1.1.1, and causes
a "This progam has performed an illegal operation
and will be shut down." message under mingw32.

I get these errors if the executable was compiled
and run under 95, or if the executable was compiled
under NT but run under 95.  (The same executable -- 
compiled under either 95 or NT -- runs fine under
NT.)

As a note, it compiles and runs fine with MSVC6.

Any thoughts as to what could be causing the problem?
Is there anything that is known to break 95 while
working under NT?

Thanks,
 - Jon

(Yes, I probably should try to get the whole source
available, but I'm in the process of trying to clean
it up for public distribution, so it may be awhile...)


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