Message-ID: red-35-msg951211061027MTP[01.51.00]000000ad-32552
I do not agree that defects in the headers that ship with the
"Windows NT Resource Kit" make the POSIX subsystem
feature of Windows NT into a joke or that such defects are
evidence of a "broken POSIX subsystem".
If the defect you mention has not been remedied, I would
agree that this can be taken as something less than full
support for POSIX-targeted development for the platform.
Whether the support that come with or in the kit is proper or
not is arguable. Given its price and nearly free availability,
I don't see that the level of support offered is improper. (I
have not yet read the warranty that may come with the kit.)
But, given that you are a customer, (I presume), I hope your
complaint gets to the right place and gets resolved. In my
mind, it sounds like a defect worthy of repair.
I don't know how to respond to your "adequate support"
issue. The support you got for your problem would appear
to be inadequate for your situation. This does not help me
to believe or disbelieve that Microsoft generally provides
support for the resource kit(s) that is adequate. (If you hold
that any failure to provide a solution makes the effort as a
whole inadequate, that greatly simplifies the issue.)
I've ported some simple tools to the POSIX subsystem,
with very little trouble, and ported a couple complex tools to
Unix systems with a lot of trouble, (FSF's Emacs to Intergraph
CLIX, for example), so my estimation of pain was based on
relevant experience. However, your defect report makes me
think there *might* be more pain than I guessed. If I were not
so busy, and if Perl[Win32] did not do what I need so well, I
would be sorely tempted to try it.
Perhaps one of the fork() lovers out there will have a go at
it. It should be quite painless to try. If it won't actually work,
and the problem is that a supported POSIX call cannot be
made to compile, link and function properly, I'll make sure
that such a defect gets to the right people to get fixed. (But
I cannot promise to fix it, since I do not work in the Windows
NT group, now.) And of course, what I mean by "supported
POSIX call" is one into the API that is a necessary part of the
claim of POSIX compliance made for Windows NT.
BTW, Bob, could you send me the atexit() call that compiles
and links under gcc but does not using just the resource kit?
I'd like to see if it's fixed, and possibly report a bug against it.
----------
| From: Bob Kline <bkline@cortex.nlm.nih.gov>
| To: Larry Brasfield
| Cc: <perl-win32@mail.hip.com>; <tchrist@mox.perl.com>
| Subject: Re: Multi-threaded server
| Date: Saturday, December 09, 1995 9:38AM
|
|
| On Fri, 8 Dec 1995, Larry Brasfield wrote:
|
| ....
| > Perhaps this is too hasty, since I'm assuming he knows fork() is
| > available to POSIX processes. (That is, processes which interact
| > with the Windows NT kernel via the POSIX subsystem.) If he was
| > ignorant of this fact, perhaps his polemics have some basis, in his
| > own world-view. But I must presume that he is not spouting off on
| > something he knows so little about.
| ....
| > If you want a POSIX version of Perl, adjust the makefile and build
| > it. You'll find that it's pretty painless.
|
| I'm afraid that this part of your arguments is based on an invalid
| assumption, that is, that Microsoft has provided adequate support for
| even the small subset of POSIX compliance which it claims.
| Unfortunately, this is not the case. I haven't tried creating a POSIX
| version of Perl, but I did make an attempt to port groff to NT using the
| POSIX subsystem, and the attempt failed miserably. For example, any
| program which called atexit failed to link, because some of the global
| symbols required for that call had been omitted from the POSIX library.
| I did get an admission from Microsoft that it was broken, but I never did
| get a fix. When the Cygnus port of gcc for NT became available, the port
| took a couple of days, which was pretty satisfying, having tried on and
| off for more than 6 months to do the port with the broken POSIX
| subsystem. Frankly, I think Microsoft's credibility with NT will rise
| significantly when it admits that the POSIX subsystem is a joke which it
| never intends to support properly.
|
| Bob Kline
|
I do not agree that defects in the headers that ship with the
"Windows NT Resource Kit" make the POSIX subsystem
feature of Windows NT into a joke or that such defects are
evidence of a "broken POSIX subsystem".
If the defect you mention has not been remedied, I would
agree that this can be taken as something less than full
support for POSIX-targeted development for the platform.
Whether the support that come with or in the kit is proper or
not is arguable. Given its price and nearly free availability,
I don't see that the level of support offered is improper. (I
have not yet read the warranty that may come with the kit.)
But, given that you are a customer, (I presume), I hope your
complaint gets to the right place and gets resolved. In my
mind, it sounds like a defect worthy of repair.
I don't know how to respond to your "adequate support"
issue. The support you got for your problem would appear
to be inadequate for your situation. This does not help me
to believe or disbelieve that Microsoft generally provides
support for the resource kit(s) that is adequate. (If you hold
that any failure to provide a solution makes the effort as a
whole inadequate, that greatly simplifies the issue.)
I've ported some simple tools to the POSIX subsystem,
with very little trouble, and ported a couple complex tools to
Unix systems with a lot of trouble, (FSF's Emacs to Intergraph
CLIX, for example), so my estimation of pain was based on
relevant experience. However, your defect report makes me
think there *might* be more pain than I guessed. If I were not
so busy, and if Perl[Win32] did not do what I need so well, I
would be sorely tempted to try it.
Perhaps one of the fork() lovers out there will have a go at
it. It should be quite painless to try. If it won't actually work,
and the problem is that a supported POSIX call cannot be
made to compile, link and function properly, I'll make sure
that such a defect gets to the right people to get fixed. (But
I cannot promise to fix it, since I do not work in the Windows
NT group, now.) And of course, what I mean by "supported
POSIX call" is one into the API that is a necessary part of the
claim of POSIX compliance made for Windows NT.
BTW, Bob, could you send me the atexit() call that compiles
and links under gcc but does not using just the resource kit?
I'd like to see if it's fixed, and possibly report a bug against it.
----------
| From: Bob Kline <bkline@cortex.nlm.nih.gov>
| To: Larry Brasfield
| Cc: <perl-win32@mail.hip.com>; <tchrist@mox.perl.com>
| Subject: Re: Multi-threaded server
| Date: Saturday, December 09, 1995 9:38AM
|
|
| On Fri, 8 Dec 1995, Larry Brasfield wrote:
|
| ....
| > Perhaps this is too hasty, since I'm assuming he knows fork() is
| > available to POSIX processes. (That is, processes which interact
| > with the Windows NT kernel via the POSIX subsystem.) If he was
| > ignorant of this fact, perhaps his polemics have some basis, in his
| > own world-view. But I must presume that he is not spouting off on
| > something he knows so little about.
| ....
| > If you want a POSIX version of Perl, adjust the makefile and build
| > it. You'll find that it's pretty painless.
|
| I'm afraid that this part of your arguments is based on an invalid
| assumption, that is, that Microsoft has provided adequate support for
| even the small subset of POSIX compliance which it claims.
| Unfortunately, this is not the case. I haven't tried creating a POSIX
| version of Perl, but I did make an attempt to port groff to NT using the
| POSIX subsystem, and the attempt failed miserably. For example, any
| program which called atexit failed to link, because some of the global
| symbols required for that call had been omitted from the POSIX library.
| I did get an admission from Microsoft that it was broken, but I never did
| get a fix. When the Cygnus port of gcc for NT became available, the port
| took a couple of days, which was pretty satisfying, having tried on and
| off for more than 6 months to do the port with the broken POSIX
| subsystem. Frankly, I think Microsoft's credibility with NT will rise
| significantly when it admits that the POSIX subsystem is a joke which it
| never intends to support properly.
|
| Bob Kline
|