Hello.

I have a windows NT server (Nec PowerMate VT, 128Mo 
RAM) with quite a lot of products installed. During 
the installation, I always get a "corrupt installation 
detected" pop-up, whereas the files can correctly be 
extracted from the archive by winzip 8 or winrar 
2.80b2.

The products installed are
- Business Bridge V2.3.16 & V3.0.6 servers (SYSTAR)
- MySQL 3.23.32 server (+ODBC driver)
- Naviscope
- I.I.S. 4.0
- Winzip 8 french
- WinRAR 2.80b2
- Cygwin 1.1
- MS SQL/Server 6.5 client (+ODBC driver)
- MouseWare 9.00.99
- Office 97
- Oracle 7.3.4 & 8i clients (+ODBC drivers)
- UltraEdit 8
- TNG Remote Control Option

I'd like to know if there are any incompatibilities 
between the installer and any of these products, and 
I'd like to be given a manual procedure that would 
allow me to install python on my workstation.

Moreover, seeing that:
- I am not the only one having installation problems 
on windows
- I have had problems every time I have tried to 
install Python 2.1 on windows (both NT and 2000)
- The Python windows installer is a 16 bit technology 
that is not supported anymore if I can believe what 
tim_one told me (see request ID 416824)
My conclusion is that we all should consider using 
another windows installer.

Regards,
  Xavier

Logged In: YES 
user_id=31435

We are having more reports of installer problems under 2.1 
than under previous releases, but I can still count the 
total on one hand.  Since we don't have the installer 
source code, and reports are still so rare (notwithstanding 
your particular bad luck) there's not much we can do to 
track them down.

Did you use the MD5 checksum program to verify your binary 
is valid?  Go to http://www.python.org/2.1/, go down to 
the "MD5 checksums" section, and click on the "MD5" part of 
the section heading to get a Python program for computing 
the MD5 digest.

In *almost* all past cases of "corrupt installation 
detected" messages, further investigation has revealed that 
the binary *was* corrupt.  The MD5 digest is a very 
powerful check on that (much stronger than a CRC):  if the 
checksum you get doesn't match the one on the web page 
above, your installer is definitely corrupt.  If it 
matches, the chance that it's corrupt anyway is too small 
to entertain.  So do that.  That you've had *repeated* 
installation failures is unprecedented in Python's history, 
so my top guess has to be you have a flaky binary.

There is only one other cause I've ever heard of for 
a "corrupt installation detected" message under NT:  the 
user was logged in to a Restricted account when they tried 
to install.  In that case, the Wise installer apparently 
can't even get at the system components it needs to compute 
its own internal checksums, and interprets that failure as 
a bad checksum.

So, what were you logged in as?  If the MD5 digest matches, 
and you're logged in to an Administrator account when it 
fails, then you've got a problem never reported before.  Do 
you?

As to using another Windows installer, I'm in favor of 
that, but so are you <wink>:  who's going to do the work?  
I can tell you I don't have time for it.  BTW, have you 
tried ActiveState's Python installer for Windows?  That's 
built with entirely different, and up-to-date, technology.  
Maybe it will work better for you.  But the terms of the 
ActiveState license don't allow us to use it too (fine by 
me -- they paid for it).

As to believing me, you don't have to:  ask Wise 
Solutions.  It's their product, version 5.0a.  You can't 
even find it mentioned on their web site anymore -- it's 
that old!  They're on version 8 now, and even their 
newsgroups only back to version 6.

Logged In: YES 
user_id=198402

OK... Here are the news: I have installed ActiveState 
version without problems. I had to: how the hell could I 
compute a MD5 digest of the Python installer with a Python 
script, if I can't have Python installed, hm?

Anyway the digest proved to be perfectly correct. After 
all, I tried downloading the file for 2 different sources, 
and this was not a copy of the file I used on Win2k (see 
request ID 416824). But computing this digest allowed me to 
find the origin of the problem. I realized that I had saved 
the installer on a remote drive (windows shared directory 
on our file server), when I had to type "V:Python-21.exe" :)

My guess is that the SYSTEM account must have the right to 
read the directory containing the installer, and it was not 
the case. Once copied locally, it seems to work (I didn't 
go through the entire process, though, as I already had 
Python installed before).

I don't know if using the msi format is free or not. 
Another way for M$ to make money with a new mandatory 
installation format? If it's free, then it's the obvious 
choice for a future version of the windows installer. I'll 
try to gather some informations about it.

And "as to believing you", I must apologize to you: this 
sentence was a dumb litteral transcription of a french 
idiom which doesn't have exactly the sense it seems to 
have. This was just a way to tell readers that this 
information did came from you, so having a look at what you 
said may be of interest. No harm intended, really.

Logged In: YES 
user_id=198402

Oops... Another mistake from a french speaking english: I 
said 'sense' but I meant 'meaning' in the last paragraph.

Logged In: YES 
user_id=31435

Closed via "Won't Fix" -- there's really nothing we can do 
about Wise 5.0a-generated installers.

Starting with 2.2a2, we're using Wise 8.14 to produce the 
Windows installer.  In my limited testing so far, it works 
much better on NT/Win2K than the old installer (which was 
produced by a version of Wise that came out when Win95 was 
brand new -- seems a miracle it ever worked at all under 
Win2K!).

BTW, every time I suggest someone compute the MD5 checksum, 
their reply contains two things:  (a) an outraged assertion 
that they can't do it since they can't install Python; and, 
(b) the news that they managed to compute it anyway.  
Apparently I give users more credit than they give 
themselves <wink>.