[OpenAFS] NOTE: Installer Bug OpenAFS for Windows 1.2.2 b

Daniel Schmidt dfs17@cornell.edu
Mon, 7 Oct 2002 02:43:18 -0400


Hi folks.

I recently installed OpenAFS for Windows 1.2.2b (as DLed from
www.openafs.org) on my Windows 2000 SP3 machine (hence I got the NT release;
dunno if the 9X release has the same issues).  I chose the default
installation, Client, Control Center, and Docs, but no Server, and found an
_extremely_ obnoxious bug:

When this program installs itself, it edits the PATH environment variable
and makes multiple registry entries, including some uninstall entries.

_Every single entry_ referring to a needed DLL (including the uninstall
DLLs) was miswritten - Not to mention the additions to the PATH statement.

They _all_ pointed to "subdirectories of C:\Program Files\IBM\..." or
something like that, a directory which OpenAFS did not create, much less
install anything into - OpenAFS installed all files in %WINNT%\, but did not
know they were there.

The results were predictable.  Immediate errors on startup about being
unable to load the appropriate library, and my complete inability to
uninstall any part of the software.

Having figured this out after a fair amount of swearing, I edited the
appropriate registry entries and verified that it was indeed functional.  As
a result I'm now somewhat less unhappy - But this _really_ needs to be
fixed.  It looks like, between the docs (IBM logos everywhere, does IBM
mind?) and the registry entries, a lot of artifacts from the days when AFS
was an IBM product still exist.

Funny thing is, the 1.2.6 "unofficial" client-only install the kind folks at
Cornell sent me a link to works just fine, and doesn't do this - Too bad
it's not available at www.openafs.org :(  The link is here BTW, in case you
want to put it on the OpenAFS site:

http://www.unc.edu/~sdw/OpenAFS_Client_126.exe

I know that no one is really actively maintaining the OpenAFS distribution
for Windows, and I don't expect that to change just 'cause of me - But since
this is so extremely annoying (and since it should be so easy to correct),
if someone who has access to these files could fix this, you'd be doing a
big favor to anyone else who might download it.

Thanks!!

PS - I'm not subscribed to the OpenAFS info mailing list, and I swear I will
trouble you no longer; I just figured someone might benefit from this
knowledge.  If you need to get in touch, feel free to drop me an e-mail.

PPS - Derrick J Brashear - Do you remember me?  CMU, '94, I lived on your
hall!  1st floor 'Schlag, what a hole that was...  Small world...  (Hi
Mitch! :)

--
Daniel Schmidt
dfs17@cornell.edu
--------------------------------------------------
Department of Materials Science and Engineering
Cornell University
--------------------------------------------------
[Opinions expressed are my own exclusively]