[R] R 2.10 under Windows XP: glitches and successes

[Duncan Murdoch]

On 29/10/2009 9:26 AM, James W. MacDonald wrote:
> Hi Duncan,
> 
> Duncan Murdoch wrote:
>> On 28/10/2009 5:54 PM, Ross Boylan wrote:
>>> I just installed 2.10 on XP; ESS is my primary interface.
>>>
>>> I seem to be able to access all the help files; under 2.8.1 I was having
>>> seemingly random problems accessing some help topics (including one day
>>> I could and a few days later I couldn't, and vice-versa).  That's good.
>>>
>>> I noticed a few glitches.
>>>
>>> First, the installer asks if you want text, CHM, or HTML help, but
>>> offers only text or HTML.  I understand CHM has been removed; I guess
>>> this is just an old message.
>> Yes, already fixed in the patched version.
>>
>>> Second, install.packages("intervals") produced the usual line about
>>> selecting a mirror, but no selection list popped up.  It just sat there
>>> until I interrupted the session and reentered the command with a
>>> pre-specified repos.  I selected custom install but took the defaults; I
>>> noticed there was some optional tk/tcl stuff (which I did not select);
>>> I'm guessing it's required for these selections.  Perhaps it should be
>>> installed by default?
>> That's something I haven't seen before.  I do get the mirror dialog, and 
>> it doesn't use tcl/tk, so I'd appreciate hearing if you can spot any 
>> pattern here (including the pattern "I never get it").  (The install 
>> option is about installing the tcl/tk help files:  mainly useful if 
>> you're developing with tcl/tk, not necessary for most users.  The Tcl/tk 
>> code itself is always installed.)  It could be that R is being blocked 
>> by your anti-virus software; what it's trying to do is to read a list of 
>> mirrors from
> 
> This is a longstanding problem for R under ess/emacs on Windows (and 
> maybe Xemacs as well, but I don't know for sure).

Okay, I'll leave it to those guys to debug.  If R is doing something it 
shouldn't, let me know what, and I'll try to fix it.  (I'd guess the 
problem is that Emacs makes assumptions about the OS that just aren't 
true in Windows, but it might be a bug we could work around.)

Duncan Murdoch

> 
> This problem manifests itself in install.packages(), but the real 
> problem comes from the line
> 
>   if (.Platform$OS.type == "windows" | .Platform$GUI == "AQUA")
>          return(.Internal(select.list(list, preselect, multiple,
>              title)))
> 
> in select.list(). If you run R under ess/emacs and do something as 
> simple as
> 
> select.list(letters)
> 
> it will freeze and you have to use C-g, C-c C-c to 'unfreeze'. 
> Paradoxically, after doing so, select.list() will then function correctly.
> 
> You can get around this problem by first loading the tcltk package, if 
> that is of any help.
> 
> This problem only occurs under emacs/ess; if you run R from a DOS prompt 
> or via the GUI then select.list() works correctly the first time.
> 
> Best,
> 
> Jim
> 
> 
>> http://cran.r-project.org/CRAN_mirrors.csv
>>
>> using the code in getCRANmirrors().  If that download fails or is 
>> delayed forever, you won't get the selection list.
>>
>>> Third, every time one asks for help on a new topic there is a
>>> considerable delay (seemed like forever, probably c 15 seconds) before
>>> the help comes up.  My first thought was that the on-demand help was
>>> being assembled, but I didn't see much CPU use.  It might be latency in
>>> talking to MSIE, and this latency might even be from my (Sophos)
>>> antivirus software, which kept warning me about a program suspiciously
>>> launching MSIE.  If you ask for help you already requested there is no
>>> delay (good) and it does not focus on the tab with the relevant help
>>> (which strikes me as the ideal behavior, though it may be hard to do).
>> When you ask for help the first time, it starts the internal server, but 
>> that should be fast.  Then it passes the URL to Windows to open, and 
>> Windows needs to open the browser to show it, which tends to be slow. 
>> When you ask subsequently, it will send the request to Windows again, 
>> but if the browser is open, that should be quick.  I'd guess this is 
>> your antivirus checking out the URL; when I do it (with no antivirus 
>> checking on such things) it's pretty slow on the first occurence 
>> (because it needs to start Firefox), then pretty fast after that.
>>
>>> Fourth, I read the previous comments about how more beta testers would
>>> be useful :)
>> Yep.
>>
>> Duncan Murdoch
>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>> PLEASE do read the posting guide 
>>> http://www.R-project.org/posting-guide.html
>>> and provide commented, minimal, self-contained, reproducible code.
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide 
>> http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>