Skip to content

macos 26 considerations

7 messages · biii+rsigm@c m@iii@g oii 8p@wexpress@com, Bill Evans, Gábor Csárdi +1 more

#
I?ve been holding off upgrading to MacOS 26 for many reasons, most 
related to yet-unsubstantiated concerns about R. Because it?s a new ABI, 
I suspect the process would be similar to a major/minor release in R: 
update the compile chain then re-install all packages. Is there a strong 
recommendation for or against this upgrade? Are there safe or at least 
less-risky pathways for upgrading the OS? Is R-devel regularly tested 
against 26? Are there known unresolved problems, such as ?package xyz 
does not currently work??

FTR: I?m currently on macos 15.7.4 using R-4.5.2, and there is nothing 
inherently broken. I don?t have a /need/ to upgrade to 26 atm but fear 
at some point that some tool I use (be it Parallels or something else) 
may eventually require 26 for some or all of the desired functionality. 
I?m just planning ahead (and holding back two colleagues, also on 
macos-15.7).

Thank you!
Bill

​
#
On 2026-02-22 10:40 a.m., Bill Evans via R-SIG-Mac wrote:
I upgraded in the fall, and found that XQuartz didn't work, and its 
maintainer had stopped working on it.  Since I was planning some work on 
rgl that needed XQuartz, I downgraded and I'm still on Sequoia 15.7.2.

You should be aware that downgrading isn't easy:  you need to create a 
bootable external disk, reformat your main drive and restore from a 
backup.  I did it and didn't lose anything, but it wasted a day or so, 
during which I wondered whether my backups were really as complete as I 
hoped they were.

Duncan Murdoch
#
I am 26.2 and everything works fine, no need to reinstall anything,
except maybe xquartz, which sometimes needs a reinstall after an
update. But xquartz works fine as well.

Gabor

On Sun, Feb 22, 2026 at 4:40?PM Bill Evans via R-SIG-Mac
<r-sig-mac at r-project.org> wrote:
#
Thank you Prof Murdoch, that?s the context I was looking for. When 
xquartz did not work for you in Tahoe, did you try homebrew or macports? 
?xorg-server? and ?xinit? packages? The pair is a documented alternative 
to xquartz, and if I believe what I read, those are supposed to work on 
tahoe.

Some follow-on discussions for the crowd:

 1.

    Does there already exist or can we make an FAQ or caution-page for
    upgrading to tahoe? I suspect I?m not the only one who has
    considered taking the upgrade, and since downgrading is not easy
    (likely ?not possible? for some), I think having a reference to this
    on |mac.r-project.org| would be a ?Really Good Thing (tm)?.

 2.

    I believe XQuartz is for displaying the plot canvas when not using
    RStudio or Positron. If we can?t have xquartz, what else are we
    losing? As an incomplete alternative, for many plots one could use
    |httpgd| (if/when they fix some bugs) or other ?immediate?
    file-based rendering, though I think this does not work for |rgl|
    (perhaps others).

 3.

    The webpage says /?R 4.6.0 CRAN arm64 builds use macOS 14?/, which
    further concerns me wrt compatibility and package availability. I
    know it?s a complicated topic, perhaps it?s too soon to say with
    authority that tahoe is going to be safe for R and for compiled
    binary packages.

    This is likely related to Xcode and the MacOS SDK; I believe Tahoe
    ?really wants? Xcode 26 and the 26 SDK, and I believe (no refs atm)
    that older versions can be problematic on Tahoe. I understand
    backward compatibility is always a concern, perhaps that is why the
    dev build is targeting macos 14, xcode 14.3, and 14.4 SDK. There are
    a lot of talented people working tirelessly (and often thanklessly)
    on this process, the more I learn the more I am thoroughly grateful
    for their knowledge and efforts.

    Is there yet a clear pathway for advancing to more up-to-date
    versions? Or is it still murky and peppering the devs with questions
    like this is just accelerating their ?early retirement??

 4.

    Are all of my xcode/sdk questions above ?moot? when using the
    RStudio or Positron IDEs? They are not (afaict) using x11 to render
    their plots, though I?m not sure how they show rgl-based plots.

Thanks all, I appreciate your time.
Bill
On 2026-02-22 08:50, Duncan Murdoch wrote:

            
&#8203;
#
Thank you Prof Murdoch, that?s the context I was looking for. When 
xquartz did not work for you in Tahoe, did you try homebrew or macports? 
?xorg-server? and ?xinit? packages? The pair is a documented alternative 
to xquartz, and if I believe what I read, those are supposed to work on 
tahoe.
On 2026-02-22 09:15, G?bor Cs?rdi wrote:

            
Very interesting! I?m curious about the difference in your experience 
and Prof Murdoch?s, perhaps the rgl-dev work is more nuanced than you?ve 
tested in your use of xquartz? I wonder if xorg-server would be any 
better (since it is quite a bit newer) for that.

Thanks all, I appreciate your time.

Bill
On 2026-02-22 08:50, Duncan Murdoch wrote:

            
&#8203;
#
On Sun, Feb 22, 2026 at 6:25?PM Bill Evans via R-SIG-Mac
<r-sig-mac at r-project.org> wrote:
[...]
I don't use rgl or x11, so I indeed only ran `x11(); plot(1:10)` for a
minimal test, and opengl indeed does not seem to work if I try
library(rgl).

If I just install xorg-server and xinit, then R does not find the X
server when loading the R_X11 dll.

G.
#
On 2026-02-22 12:15 p.m., Bill Evans wrote:
No, I didn't try those. I was on a very tight schedule, and I didn't 
have time to experiment.  But from what Gabor wrote, maybe Xquartz is 
working again with the latest Tahoe.

For what it's worth, I did a quick search today and found this advice 
against Tahoe for some neuroimaging people:

  
https://discuss.afni.nimh.nih.gov/t/re-macos-tahoe-and-xquartz-issues-for-afni-users/9227

As of Jan 21 they were still advising against the upgrade, though some 
of their problems had been addressed.

Duncan