[oops, forgot to copy r-help, which could be rather important to do, so
resending]
There's a difference between "getting away with in the past" and "safe to
do"....
I just had another conversation (off-line) with someone getting bitten by the
change to md5sum(), and I'm 99% sure it happened because copying the old
library overwrote the new tools package with the old one, an updated package
tried to call tools::md5sum(bytes=....), and things blew up.
So even though people have apparently gotten away with copying the old folder
in the past, they probably actually created a bastardized installation, where
some system packages was actually from a previous version (which of course is
inconsequential if and only if they weren't actually changed by the update...).
- pd
> On 24 Apr 2025, at 14:31 , Evan Cooch <evan.co...@gmail.com> wrote:
>
> Good advice, but that approach (as you describe it) is not something I (or
> anyone I know) has needed to follow for past updates to R, and apparently
> is/was only a consideration for BH. As per OP, every other package updated
> just fine "the old way" (which you correctly describe as what most of us do
> -- copy old /library folder to 'new' /library folder, and simply update
> everything).
>
> On 4/24/2025 7:32 AM, peter dalgaard wrote:
>> I'm puzzled. It doesn't seem to take long on MacOS, neither under 4.3.x. or
>> 4.5.0.
>>
>> The tarballs for BH are like 14 MB and the Windows .zip is 21 MB, which is
>> large, but not that large?
>>
>> I'm somewhat suspicious about the "package updating" techniques that some
>> people seeme to be using. It sounds like you are copying package directories
>> from an older version, then using update.packages(). A number of things
>> could go wrong if you do that across an API change.
>>
>> At the very least, you need update.packages(checkBuilt=TRUE) (and probably
>> you also want ask=FALSE), but I would prefer a safer strategy of just
>> installing what that you need into a fresh install of R. I.e. something like
>> this:
>>
>> In the old R do
>>
>>> pkglist <- rownames(subset(as.data.frame(installed.packages()), !(Priority
>>> %in% c("base","recommended"))))
>>> save(pkglist, file="~/tmp/4.4pkglist")
>>>
>> Install new version, then
>>
>>> load("~/tmp/4.4pkglist")
>>> system.time(install.packages(pkglist))
>>>
>> --- Please select a CRAN mirror for use in this session ---
>> trying URL '
>> https://cloud.r-project.org/bin/macosx/big-sur-x86_64/contrib/4.5/abind_1.4-8.tgz
>> '
>> trying URL '
>> https://cloud.r-project.org/bin/macosx/big-sur-x86_64/contrib/4.5/ada_2.0-5.tgz
>> '
>> tryi.......
>> ......ng URL '
>> https://cloud.r-project.org/bin/macosx/big-sur-x86_64/contrib/4.5/zoo_1.8-14.tgz
>> '
>>
>> The downloaded binary packages are in
>>
>> /var/folders/h0/hzzhnnfd1gx7399sx_cwlhpw0000gn/T//RtmpQDl3Tj/downloaded_packages
>> user system elapsed
>> 4.574 5.235 33.213
>> Warning message:
>> package ‘fEcofin’ is not available for this version of R
>>
>> A version of this package for your version of R might be available elsewhere,
>> see the ideas at
>>
>> https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages
>>
>>
>>> length(pkglist)
>>>
>> [1] 252
>>
>> (most of the time was spent waiting for XQuartz and Tcl/Tk to fire up so
>> that I could choose the CRAN mirror.)
>>
>>
>> Actually, BH was not on the list, but
>>
>>
>>> system.time(install.packages("BH"))
>>>
>> trying URL
>> 'https://cloud.r-project.org/bin/macosx/big-sur-x86_64/contrib/4.5/BH_1.87.0-1.tgz
>> '
>> Content type 'application/x-gzip' length 13900146 bytes (13.3 MB)
>> ==================================================
>> downloaded 13.3 MB
>>
>>
>> The downloaded binary packages are in
>>
>> /var/folders/h0/hzzhnnfd1gx7399sx_cwlhpw0000gn/T//RtmpQDl3Tj/downloaded_packages
>> user system elapsed
>> 1.054 2.340 5.091
>>
>> -pd
>>
>>
>>
>>> On 23 Apr 2025, at 23:33 , Evan Cooch <evan.co...@gmail.com>
>>> wrote:
>>>
>>> So, I decided to update from 4.4.3 -> 4.5.0. Updating all packages in
>>> one shot (selecting all of the offered updates) failed, on all my
>>> machines (Linux, Windows - lots of RAM, lots of CPU speed). So, tried
>>> the updates a few at a time. Didn't take me long to discover that the
>>> 'culprit' was the BH packages. For some reason, it takes a *long* time
>>> for it to download/update or download/install. Even installing from a
>>> local download took a long time. So long, that on my Windows machines,
>>> the Rgui simply stopped responding. At all. Had to kill the process
>>> manually, and try again.
>>>
>>> So, in case anyone else runs into this (I've replicated the problem on 5
>>> different machines - 3 Windows, 2 Linux)
>>>
>>> 1\ if BH is included in a slew of updates you're trying to apply all at
>>> once, it might clobber everything because its choking on BH, kill RGui
>>> on my Windows machines. Less of an issue on my Linux boxes (good old
>>> CLI), but still...
>>>
>>> 2\ even BH alone requires some patience. On a high-end machine (24
>>> threads, 64 Gb RAM, etc etc) took almost 8 minutes to download/update.
>>> Everything else (even big suckers like terra) took only seconds on said
>>> machine. But BH?
>>>
>>>
>>> [[alternative HTML version deleted]]
>>>
>>> ______________________________________________
>>>
>>> R-help@r-project.org
>>> mailing list -- To UNSUBSCRIBE and more, see
>>>
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>
>>> PLEASE do read the posting guide
>>> https://www.R-project.org/posting-guide.html
>>>
>>> and provide commented, minimal, self-contained, reproducible code.
>>>
>
--
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd....@cbs.dk Priv: pda...@gmail.com
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide https://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
