[Rd] Segfault in setMask in R 4.1
Mike Lee Williams
m|ke @end|ng |rom m|ke@p|@ce
Wed Sep 22 05:27:02 CEST 2021
I'm all set! My problem seems to have been a mixture of two things:
1. As Paul pointed out I needed to upgrade Cairo to get support for the new GE v13 (setMask, etc.). That support is apparently present in 1.5-12.2 (the current version on CRAN).
2. I was getting Cairo from https://github.com/s-u/Cairo/ rather than CRAN, and github and CRAN seem to have diverged, such that 1.5-13 (on github) doesn't contain the GE v13 change despite being a higher version than 1.5-12.2.
Simon Urbanek, the Cairo maintainer has kindly cleared this situation up for me, and there is now a 1.5-14 which I believe reconciles these two diverged sources, includes the GE v13 changes (see https://github.com/s-u/Cairo/) and fixes my segfual. I assume this will be pushed to CRAN at some point.
Thank you very much Paul, Andre and Simon for you help!
Mike
On Tue, Sep 21, 2021, at 2:24 PM, Paul Murrell wrote:
> Hi
>
> The Cairo_1.5-12.2.tar.gz sources on CRAN contain the changes. It looks
> like they have still not been merged back into the main package sources
> on R Forge (cc'ing the package maintainer as a gentle reminder).
>
> I think the Cairo package NEWS file should say 4.1.0 rather than 4.0.0
>
> Paul
>
> On 9/22/2021 9:08 AM, Mike Lee Williams wrote:
>> Thanks! AFAICT I am already using the latest version of Cairo, but these
>> links gives me a ton of leads to follow up.
>>
>> One question (which may be better directed at the maintainer of Cairo):
>> exactly which version (or commit) of Cairo implements support for the
>> new setMask API?
>>
>> CRAN still points to 1.5-12.2. The NEWS file
>> (https://www.rforge.net/Cairo/news.html
>> <https://www.rforge.net/Cairo/news.html>)
>> refers to changes implemented to support R 4.0.0 in Cairo 1.5-12, but my
>> understanding is that setMask was added in R 4.1.0.
>>
>> I see no mention of changes to support setMask or R 4.1.x in Cairo
>> 1.5-13 (not on CRAN yet, but the only release since 1.5-12 according to
>> the NEWS file). The 1.5-13 tarball at
>> https://www.rforge.net/Cairo/files/
>> <https://www.rforge.net/Cairo/files>
>> contains no mention of strings like "setMask", "releasePattern" or
>> "setPattern", which seem like they should be in there if that version
>> has implemented support for the new API.
>>
>> Mike
>>
>> On Tue, Sep 21, 2021, at 1:41 PM, Paul Murrell wrote:
>> > Hi
>> >
>> > dev->setMask() was introduced in R 4.1.0 ...
>> >
>> >
>> https://developer.r-project.org/Blog/public/2020/07/15/new-features-in-the-r-graphics-engine/
>> <https://developer.r-project.org/Blog/public/2020/07/15/new-features-in-the-r-graphics-engine>
>> >
>> >
>> https://www.stat.auckland.ac.nz/~paul/Reports/GraphicsEngine/definitions/definitions.html
>> <https://www.stat.auckland.ac.nz/~paul/Reports/GraphicsEngine/definitions/definitions.html>
>> >
>> > If your application is using its own custom graphics device, it will
>> > need updating, see ...
>> >
>> >
>> https://www.stat.auckland.ac.nz/~paul/Reports/GraphicsEngine/definitions/definitions.html#devices
>> <https://www.stat.auckland.ac.nz/~paul/Reports/GraphicsEngine/definitions/definitions.html#devices>
>> >
>> > But it looks like you may just be using the graphics device provided by
>> > the 'Cairo' package, which has already been updated. If that is the
>> > case, your problem may be solved simply be reinstalling the 'Cairo'
>> package.
>> >
>> > Paul
>> >
>> > p.s. the reason why it only happens for 'ggplot2' plots is that only
>> > 'grid' (which underlies 'ggplot2') attempts to set masks (the 'graphics'
>> > package does not).
>> >
>> > On 9/22/2021 4:52 AM, Mike Lee Williams wrote:
>> >> Thank you! Here is the output of option("devices"), which I think means
>> >> we're using Cairo:
>> >>
>> >> $device
>> >> function (width = 1280, height = 960, pointsize = 24, units = "px",
>> >> bg = "white", dpi = 160, ...)
>> >> {
>> >> devSym <- basename(tempfile(pattern = "SensePlot", tmpdir = ""))
>> >> newDev <- Cairo(width = width, height = height, pointsize = pointsize,
>> >> bg = bg, units = units, dpi = dpi, type = "raster", ...)
>> >> attr(newDev, "units") <- units
>> >> attr(newDev, "dpi") <- dpi
>> >> assign(devSym, newDev, SenseDevices)
>> >> invisible(newDev)
>> >> }
>> >> <bytecode: 0x55b08ba031c0>
>> >> <environment: namespace:Cairo>
>> >>
>> >> "Sense" is the internal name of our application. I'm not sure what
>> >> "devSym" and "newDev" are, but If there are no known issues with Cairo
>> >> itself then our own code is presumably part of the problem.
>> >>
>> >> I would like to confirm that by failing to reproduce the bug in
>> >> "vanilla" R 4.1 with a Cairo device. If anyone has any tips for how to
>> >> do that (on Linux) or any other comments I would be very grateful.
>> >>
>> >> If our code is the problem then I'm still confused why a regular plot()
>> >> works, but a ggplot2 qplot() segfaults. Maybe this is a clue!
>> >>
>> >> I will post again if I figure anything out!
>> >>
>> >> Mike
>> >>
>> >> On Tue, Sep 21, 2021, at 3:37 AM, GILLIBERT, Andre wrote:
>> >> > Hello,
>> >> >
>> >> > Which graphic device (backend) do you use?
>> >> > The setMask() function calls the internal setMask function associated
>> >> > to the graphic device.
>> >> > If the Web application uses one of the standard graphic backend, the
>> >> > bug may come from the R core. If it uses a custom graphic backend, the
>> >> > bug may be in the backend.
>> >> >
>> >> > --
>> >> > Sincerely
>> >> > André GILLIBERT
>> >> >
>> >> > ATTENTION: Cet e-mail provient d’une adresse mail extérieure au CHU de
>> >> > Rouen. Ne cliquez pas sur les liens ou n'ouvrez pas les pièces jointes
>> >> > à moins de connaître l'expéditeur et de savoir que le contenu est sûr.
>> >> > En cas de doute, transférer le mail à « DSI, Sécurité » pour analyse.
>> >> > Merci de votre vigilance
>> >> >
>> >> >
>> >> > I have inherited a build of R. We compile R from source and then use
>> >> a custom
>> >> > web application to allow users to enter R statements and render the
>> >> output (it's
>> >> > kind of like RStudio although the UX is quite different).
>> >> >
>> >> > Things were going fine until I tried to upgrade to R 4.1.x. The build
>> >> succeeds,
>> >> > but I get the following segfault when I make qplot (ggplot2) calls:
>> >> >
>> >> > *** caught segfault ***
>> >> > address (nil), cause 'memory not mapped'
>> >> >
>> >> > Traceback:
>> >> > 1: .setMask(NULL, NULL)
>> >> > 2: resolveMask.NULL(NULL)
>> >> > 3: (function (path) { UseMethod("resolveMask")})(NULL)
>> >> > 4: grid.newpage()
>> >> > 5: print.ggplot(x)
>> >> > 6: (function (x, ...) UseMethod("print"))(x)
>> >> >
>> >> > I am not an R developer, so I don't really know how to read this. I
>> >> am wondering
>> >> > if this is a known issue or if anyone has any suggestions. Here's
>> >> some things
>> >> > I've tested:
>> >> >
>> >> > - I get the same segfault in R 4.1.0 and R 4.1.1. I do not get this
>> >> error in R
>> >> > 4.0.4 or R 4.0.5.
>> >> >
>> >> > - The problem appears to be specific to the (graphics?) features of R
>> >> that
>> >> > ggplot2 uses. I do not get a segfault if I do a generic
>> `plot(c(1,2,3))`.
>> >> >
>> >> > - I have tried versions of ggplot2 3.3.3 and 3.3.5.
>> >> >
>> >> > - The traceback points to files that were introduced in this commit
>> >> >
>> >> >
>> >>
>> https://github.com/wch/r-source/commit/16755bcddffe0cb4238d8a4979387d92b93a8324#diff-5c63f74229830cdde7886a50bf06dafcdf1d5b3d42cfa06b26814876e58c5ab0
>> <https://github.com/wch/r-source/commit/16755bcddffe0cb4238d8a4979387d92b93a8324#diff-5c63f74229830cdde7886a50bf06dafcdf1d5b3d42cfa06b26814876e58c5ab0>
>>
>> >>
>> <https://github.com/wch/r-source/commit/16755bcddffe0cb4238d8a4979387d92b93a8324#diff-5c63f74229830cdde7886a50bf06dafcdf1d5b3d42cfa06b26814876e58c5ab0
>> <https://github.com/wch/r-source/commit/16755bcddffe0cb4238d8a4979387d92b93a8324#diff-5c63f74229830cdde7886a50bf06dafcdf1d5b3d42cfa06b26814876e58c5ab0>>
>> >> >
>> >> > I am not an R user or R developer so I'm a bit stuck here. I would be
>> >> happy to
>> >> > give the output of any commands. If anyone has any suggestions then
>> >> please let
>> >> > me know!
>> >> >
>> >> > Thanks!
>> >> > Mike Lee Williams
>> >> >
>> >> > ______________________________________________
>> >> > R-devel using r-project.org mailing list
>> >> > https://stat.ethz.ch/mailman/listinfo/r-devel
>> <https://stat.ethz.ch/mailman/listinfo/r-devel>
>>
>> >> <https://stat.ethz.ch/mailman/listinfo/r-devel
>> <https://stat.ethz.ch/mailman/listinfo/r-devel>>
>> >>
>> >> ______________________________________________
>> >> R-devel using r-project.org mailing list
>> >> https://stat.ethz.ch/mailman/listinfo/r-devel
>> <https://stat.ethz.ch/mailman/listinfo/r-devel>
>>
>> >> <https://stat.ethz.ch/mailman/listinfo/r-devel
>> <https://stat.ethz.ch/mailman/listinfo/r-devel>>
>> >
>> > --
>> > Dr Paul Murrell
>> > Department of Statistics
>> > The University of Auckland
>> > Private Bag 92019
>> > Auckland
>> > New Zealand
>> > 64 9 3737599 x85392
>> > paul using stat.auckland.ac.nz
>> > http://www.stat.auckland.ac.nz/~paul/
>> <http://www.stat.auckland.ac.nz/~paul>
>
> --
> Dr Paul Murrell
> Department of Statistics
> The University of Auckland
> Private Bag 92019
> Auckland
> New Zealand
> 64 9 3737599 x85392
> paul using stat.auckland.ac.nz
> http://www.stat.auckland.ac.nz/~paul/
More information about the R-devel
mailing list