Skip site navigation (1) Skip section navigation (2)

Re: BUG #5667: install failure

From: Craig Ringer <craig(at)postnewspapers(dot)com(dot)au>
To: Roedy Green <roedyg(at)mindprod(dot)com>
Cc: pgsql-bugs <pgsql-bugs(at)postgresql(dot)org>
Subject: Re: BUG #5667: install failure
Date: 2010-09-28 08:04:07
Message-ID: (view raw, whole thread or download thread mbox)
Lists: pgsql-bugs
On 28/09/10 15:51, Roedy Green wrote:
>>>>  It'd also help to know:
>>>>  - If you run the installer from the command line as:
>>>>      postgresql-9.0.0-1-windows.exe --install_runtimes 0
>>>>    does it install without error?
> I ran this using the standard Windows cmd processor.  I got exactly the
> same behaviour as with JPSoft Take Command.

I assume you mean "the same behaviour as the above command run with Take
Command" - because the initial issue you reported was with the runtimes
failing to install, and the above command shouldn't have attempted to
install the runtimes.

Did you try running the installer in extract-only mode, and then running
the runtime install script manually? This would be a useful diagnostic
test. I explained it in my previous mail, but you said you hadn't tried
it yet because you thought it only had to be done when the installer
succeeded. The point is that you run the installer in extract-only mode,
which should succeed if the problem is in fact with the scripts, then
you run the scripts manually. I'll quote what I said earlier:


- Does the VC++ runtime installer executes successfully when run stand-
  alone? You can unpack the VC++ runtime installer by running the Pg
  installer as:

  postgresql-9.0.0-1-windows.exe --extract-only 1 --prefix "%TEMP%\pg9"

  This should succeed, even if the installer normally fails. If it
  still fails, please mention that along with the error it reports,
  and don't try to continue with these instructions.

  If the extract-only installer run does succeed:

  Now you can open an Administrator-rights cmd.exe window by hitting
  Start, entering "command", waiting until "command prompt" appears,
  right-clicking on it and choosing "run as administrator". Into the
  cmd.exe window that appears enter:

  cd "%TEMP%\pg9\installer"

  You'll see the Visual C++ runtime installer (vcredist) and the
  vbscript used to invoke it.

  Run the vbcscript by typing "install" and pressing tab, pressing
  space, typing "vcredist" and pressing tab, then pressing enter.

  You should get a message saying vcredist is being installed, a
  progress bar, then a message saying it installed successfully.


Anyway, this is looking pretty strange. It's really hard to know what on
your machine is throwing the installer out of whack, and without knowing
what's going on it's hard to know what if anything can be done to make
the installer more robust.

Do you have any Group Policy on the machine, by any chance?

There seems to be something wrong with execution of the vbs scripts on
your machine, but it's not clear what exactly it is. I wonder how many
of the other installer-related issues are with vbs script execution too?

At this point it's beyond anything I can do without access to your
machine and/or the installer sources. I don't have either. If I were
sitting at your machine right now, I'd be firing up Process Monitor from
sysinternals and leaving it to run for a little while to build an
impression of the background activity on the machine. Then I'd fire up
the PostgreSQL installer and record the activity in Process Monitor.
After that I'd spend an extended period of rather tedious time poring
over the logs trying to see what might be going wrong.

I guess you could possibly install Process Monitor and the rest of the
sysinternals suite from:

then collect the process monitor data, export it, upload it somewhere,
and post a link on the list. I seem to remember that process monitor's
export format is MUCH less useful than it's live format, though, so I
don't know if it'll be much help.

Hopefully one of the EnterpriseDB folks will be able to look into what's
happening here, time permitting. They're doing the Windows installer as
a free service to the community after all. I'm toward the end of what I
can do in terms of diagnosis and data collection.

Craig Ringer

Tech-related writing:

In response to


pgsql-bugs by date

Next:From: Dave PageDate: 2010-09-28 08:57:08
Subject: Re: BUG #5667: install failure
Previous:From: Tom LaneDate: 2010-09-28 07:04:34
Subject: Re: BUG #5680: Failure to start: too many private dirs demanded

Privacy Policy | About PostgreSQL
Copyright © 1996-2017 The PostgreSQL Global Development Group