[R-pkg-devel] problem with regdmp finding v4.2.x under Windows

[Tomas Kalibera]

On 7/5/22 19:27, Kevin Ushey wrote:
> Hi Jonathan,
>
> I'm not certain if this is the issue you're encountering, but I've
> noticed that different versions of R may write to different locations
> in the registry.

The R installer does not write R version information to the 32-bit view 
of the registry when installing only 64-bit files. This is the same with 
previous versions of R as can be seen from the code. I've tested with 
4.1.3 - when you opt out from installing 32-bit files, version 
information won't appear in the 32-bit view. In principle, one could 
still write something to the 32-bit view, though I think this decision 
was intentional as it required an extra condition in the sources.

R 4.2.x doesn't support 32-bit files anymore, hence  for users using the 
defaults, it is R 4.2.0 when the version information no longer appears 
in the 32-bit view.

One could proceed by searching both the 32-bit and the 64-bit views when 
looking for the newest installed version, preferring 64-bit. For 
practical use on current systems, using the 64-bit view only should 
probably do.

Best,
Tomas

>
> Right now, on my machine, I see entries for R 4.1.3 installed at:
>
>      Computer\HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\R-core\R
>
> And for my installation of R 4.2.1, I see it at:
>
>      Computer\HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R
>
> Note that the registry location for R 4.2.1 is no longer part of the
> WOW6432Node.
>
> I'm not sure if this is a hiccup unique to my own local machine, but
> perhaps you're seeing something similar? A way to verify would be to
> check and compare what you see with:
>
>      reg query HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R /reg:32
>      reg query HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R /reg:64
>
> In my case, I see:
>
> C:\Users\kevin>reg query HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R /reg:32
>
> HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R
>      InstallPath    REG_SZ    C:\R\R-4.1.3
>      Current Version    REG_SZ    4.1.3
>
> C:\Users\kevin>reg query HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R /reg:64
>
> HKEY_LOCAL_MACHINE\SOFTWARE\R-core\R
>      InstallPath    REG_SZ    C:\R\R-4.2.1
>      Current Version    REG_SZ    4.2.1
>
> Best,
> Kevin
>
> On Tue, Jul 5, 2022 at 8:00 AM Uwe Ligges
> <ligges using statistik.tu-dortmund.de> wrote:
>> The installer asks you
>>
>> "Save Version in Registry" with Description "Save the R version and
>> install path in the Registry". If this is checked in the installation
>> wizard, it should work, otherwise not.
>>
>> Best,
>> Uwe Ligges
>>
>>
>> On 05.07.2022 10:15, Ivan Krylov wrote:
>>> Hi Jonathan,
>>>
>>> Since we don't have a separate "R on Windows" or "R installation
>>> and administration" mailing list, I think that this question is a
>>> better fit for R-help, not R-package-devel. Although R-devel could be
>>> an option too (if we determine that the R installer is doing something
>>> wrong), the last time R installer has been significantly changed is
>>> 2020, so the problem is likely elsewhere.
>>>
>>> On Tue, 5 Jul 2022 01:59:43 +0000
>>> Jonathan Godfrey <A.J.Godfrey using massey.ac.nz> wrote:
>>>
>>>> To do all of that efficiently, I make extensive use of batch files
>>>> (Windows user alert) and a key ingredient of those batch files is
>>>> that I avoid having to edit the R version number with each
>>>> installation.
>>> Can you show us the relevant parts of the batch files?
>>>
>>>> On my older desktop which has multiple R installations, I do not get
>>>> an "Access denied" error, quarto finds the path to the 4.2.0
>>>> installation via brute force, but regdmp via my batch files gives me
>>>> 4.1.x instead.
>>> What is the output of the following commands on the machines where you
>>> have troubles finding R-4.2.x?
>>>
>>> reg query "HKEY_LOCAL_MACHINE\Software\R-core"
>>> reg query "HKEY_CURRENT_USER\Software\R-core"
>>>
>>>>       [[alternative HTML version deleted]]
>>> This mailing list removes the HTML parts from our messages, so we only
>>> see the automatically-generated plain text part. In this case, I think
>>> that your message wasn't damaged at all, but sometimes (especially when
>>> pasting code or output examples) the results can be undesirable.
>>>
>> ______________________________________________
>> R-package-devel using r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-package-devel
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel