[wellylug] CVS woe

[Cliff Pratt]

Daniel Pittman wrote:
> Cliff Pratt <enkidu at cliffp.com> writes:
>> Daniel Pittman wrote:
>> 
>>> Ah, Cygwin, which is the single most god-awful POSIX environment 
>>> ever, bypassing even the "Unix in FORTRAN" stuff that was popular
>>>  decades ago.
>>> 
>>> You might find your life radically improved by using the
>>> "Services For Unix" enhanced POSIX layer for Windows, which
>>> doesn't suck nearly as much.
>>> 
>>> That, at least, would eliminate the Cygwin based stupidities from
>>> the system for you.
>> SFU - the acronym says it all. I still have the nightmares. The
>> "POSIX layer" AKA "Interix" was/is truly horrible.
> 
> Compared to Cygwin, or just in general?  The NFS and NIS integration
> was a bit pants, really, but the basics were good.
> 
NFS + Interix was SFU. Well, I'm talking about version 2.0, but I don't
suppose that change in later versions. (But see at the end).
> 
>> Cygwin's environment doesn't pretend to be POSIX compliant so far
>> as I am aware, does it?
> 
> I really don't know how to respond to that.  Um, yes?
> 
> I suppose you could disregard the marketing fluff they have left for 
> Google to find in the results at
> http://www.google.com.au/search?q=cygwin
> 
> Cygwin Information and Installation
> 
> Linux-like environment for Windows making it possible to port 
> software running on POSIX systems (such as Linux, BSD, and Unix 
> systems) to Windows.
> 
> in favour of their "Linux-like" claims on the front page.
> 
> At that point, though, you would be arguing that a "POSIX and stuff" 
> goal is not the same as POSIX.
> 
Oh OK.
> 
>> It's hard to choose between the two, really. SFU (I'm not sure if
>> it was the Interix component) used to regularly hose the Windows
>> file permissions.
> 
> SFU is Interix, and Interix is SFU.  That might explain the
> difference in our experiences, though:
> 
(But see at the end)

Erm, not that I remember. SFU was SFUed NFS with *parts* of Interix on
top. It wasn't a full POSIX implementation (though the full Interix 
was). You couldn't run a shell, you couldn't run scripts, and so far as 
I could tell it was only the bits of Interix that would allow a Windows 
NFS solution. We ran it on Windows NAS servers  which HP sold as NAS 
'appliances'.

I went on a search for the Interix bits and I found at least some of the 
programs, but I didn't find a shell though I did find Perl.
> 
> Interix delivers a genuine POSIX implementation on top of the NT
> kernel, running next to the Win32 layer.  That includes POSIX
> semantics in the files system.
>
Yes, but not in SFU. It was only a partial implementation of the full 
Interix. (But see at the end)
 >
> You could use it to confuse Win32 software by having case-sensitive 
> filenames ("Foo" vs "foo", in the same directory) or through having 
> different permission inheritance rules (POSIX rather than NT ACL 
> inheritance).
>
Ah, maybe the different permission inheritance rules would explain why 
all Windows file permissions would now and then disappear.

Hah! I've discovered a possible reason for our different experiences. 
The systems we had were described as NAS4000 with Windows 2K3 Server 
Appliance Edition (or similar). Please see the following link and 
especially the bit that talks about "NAS4000 w/ win2k3, Server Appliance 
Kit(SFU 2.5), and SR1 patches CD installed" and further on down about "I 
have read in other threads that removing the Server Appliance Kit and 
installing the full SFU 3.5 version resolves the issue, but the 
documentation with the NAS4000 specifically states that this should NOT 
be done??"

http://tinyurl.com/c7vlwa or
http://forums11.itrc.hp.com/service/forums/bizsupport/questionanswer.do?admit=109447626+1238839841009+28353475&threadId=709181

It looks to me that the "Server Appliance Kit (2.5)" might have been a 
crippled version of the SFU. (I thought we had SFU2.0 but could be 
wrong). What we had was definitely NOT the full Interix.

Cheers,

Cliff