Problems installing package on Windows - and eventual success . . . . . . .

[Jay Walker]

I installed this package on Win Vista but I left out mention of Vista on the subject line because I highly doubt the issues were Vista-specific.

As you will see, I might not have run into some of these issues if I had uninstalled one or more programs before attempting to install this package.

Maybe it's just as well as it seems to have exposed weakness(es?) with the installer program.

Problem #1: It appears that installer does not handle its disk management properly, including the sandbox.

Detail:

Over a year ago I installed some app that created a RailsInstaller folder right under the C:\ drive. Where did that come from ??

I don't remember doing anything ruby-centric at the time. In fact, I think that's when I experimented with Django but soon gave up

after installing it. Does anyone out there know if Django or any other package for that matter creates a folder called C:\RailsInstaller ?

I strongly suspect it was something other than ruby because my installed programs (control panel) showed an entry for RailsInstaller

(and therefore provides the option to UNinstall it), whereas now that I have this package installed and working, there is now NO entry in 

my installed programs like this. CAN SOMEONE OUT THERE CONFIRM THAT THERE ARE OTHER PACKAGE INSTALLERS, 

DJANGO OR OTHERWISE, THAT CREATE C:\RailsInstaller AND create an entry for same in installed programs (control panel) ??

Anyway, I had no clue that this ruby rails installer was going to try to create the same folder (C:\RailsInstaller), so I ran the installer and, 

interesting, it got all the way through without reporting any errors, so I thought I had success. Man, was I mistaken !  Early in the process it 

told me it was going to create C:\RailsInstaller and gave me a chance to change that path, but I didn't bother. So when I then went to check 

for rails version at the command line [ > rails -v ], nobody home. Did a bit of probing and quickly realized the dates on my C:\RailsInstaller

folder were 14 months old.  Also, when I went to look for the version of ruby I had installed  [ > ruby -v ],  it reported back with the old version 

that was installed independent of rails or anything else a couple weeks ago (1.9.x) whereas I was expecting the version that is advertised 

on the RailsInstaller web page (2.1.5 currently). Quickly concluded I needed to UNinstall ruby and RailsInstaller using installed programs 

(control panel). Did so, then attempted to again install this ruby RailsInstaller package. That's when it became apparent that a one-click install 

(yeah, right) was going to turn into an all-nighter. 

Shortly after attempting the reinstall, I got an error message that the installer program could not create C:\RailsInstaller because that folder already

existed. It was correct, because the earlier UNinstall of RailsInstaller (via control panel) did not remove this folder like it should have. So I 

removed it manually. Guess what. Same error on next reinstall attempt. HUH ?? Was this one of those funny Windows things where it was looking

at some earlier cached version of disk status or something ?  Or a registry entry that should have been deleted during the uninstall but didn't ?

I cleaned out the registry with a good app for that, but it made no difference. I couldn't figure it out. I was considering monkeying with the registry

directly (YIKES - VERY DANGEROUS !!) but luckily I tried a different approach that led me to the man behind the curtain. I searched for RailsInstaller

using the OS search tool. It revealed RailsInstaller as well as other related folders in a hidden folder called C:\VTRoot\Harddiskdrive1.  Like other installer

programs, I assume, the ruby RailsInstaller operates in a sandbox-like proxy folder(s) for nearly the entire install process until it gets to the end of the process,

then, only if a SUCCESS flag is set, only then does it copy the proxy folder(s) to the intended aforementioned destination(s) (namely, C:\RailsInstaller).

So, by deleting most of the folders in C:\VTRoot\Harddiskdrive1 (not all, because one of the folders corresponded to a different app entirely, 

so I left that one alone, and evidently, this hidden folder is known to other app installers).  There had been multiple delete folders/reinstall attempts iterations 

up to this point. But this was the beginning of the breakthrough. The 6th (or 7th ?) reinstall (with a couple of system restarts in there somewhere) subsequently

went much better, or appeared to, but I still wound up having to copy some folders from C:\VTRoot\Harddiskdrive1 to the main C:\ drive (e.g., the installer also

tries to create C:\Sites). At the end of what the installer thinks is a successful install, it launches a cmd.exe window with some info displaying things like where 

ruby is installed and where rails is installed.  I had the presence of mind to type [ > PATH ] into this shell so it would display the appended path that includes 

the new ruby and rails folders, because I invoked a new cmd shell from the OS and it did NOT include these appends, so I copied the differences into the clipboard 

and added them onto the system environment variables; perhaps a bit brute-force, but it works, because after all that, I was then able to get the expected 

response to [ > rails -v ] and  [ > ruby -v ] in a new cmd shell. Now things were looking a lot brighter and I was at least able to begin to have some success 

following the tutorial to create my first rails app. But I quickly ran into another problem detailed below.  Still and all.........PHEW !! Made a lot of progress !

 

Conclusion #1: The installer needs to be made more robust so that:

                        (1) When it tells us that the folder C:\RailsInstaller already exists (or whatever substitute folder the user selected), it is actually telling the truth.

                        (2) And in the event it properly senses that the install did not complete, it should delete the hidden folders it created in C:\VTRoot\Harddiskdrive1; 

                            otherwise we have to do that to have any hope of a subsequent successful install.

 

Problem #2:

After I got everything installed I tried to use the package to create a very basic rails app and I soon ran into what I am calling a package integration issue.

I didn't document the error at the time but it had everything to do with existing RAKE gem version not compatible with the Bundler package component.

The error message was helpful enough I was able to figure out I had to update the RAKE gem version, after which the error went away. Can't recall all the

gory details, but the following two web pages were helpful to me in resolving the matter:

http://guides.rubygems.org/rubygems-basics/#installing-gems

http://guides.rubygems.org/ssl-certificate-update/

Conclusion #2:  It appears to me that the person (or people ?) maintaining RailsInstaller for Windows could do more to keep up with changes/updates 

                         with the various components so as to avoid integration issues such as I experienced. Probably easier said than done, I appreciate, 

                         but between this and the installer issues detailed above, the alternative is that a one-click install morphs into an unpaid part-time 

                         sys-admin job for all us end users.