Windows build using Rtools libraries? #9
Comments
|
I made a couple of fixes, but most of these relate to how the Windows binaries are built. It looks like you'll need libtiff and pcre. I'm not sure why some of the win32-specific symbols in GDK are missing (like |
|
Thanks - I will take a look when I'm back in my Windows partition. It looks like linking order is a really tricky problem in mingw32, but there are potential workarounds: https://cran.r-project.org/bin/windows/base/howto-R-4.2.html |
|
By using the Unfortunately, The resulting DLL can't be loaded. In R we get To debug this error, I've run WinDbg as described in it the Rtools HOWTO. I get these errors (among lots of other output): winspool.drv seems to be in c:\Windows\System32, but I can't find netio.sys. I'll have to think about this another day... |
|
Checking using winbuild-devel (output attached) shows that the linking problem is not resolved with RTools4.3 - so the next step is to try that out. |
|
Here is the results of trying with RTools4.3 and the appended contents of The package can be installed, but the resulting DLL can't be loaded. There are the same errors observed when running Full output attached. Contents of |
|
Here is the output from |
|
What is the PATH environment variable set to in the place where you're running R? My understanding is those drivers are normally located in the "standard" system locations, e.g. C:\Windows\System32, so the only thing I can imagine is that for some reason they're not on the PATH. |
|
@kevinushey I will double-check, when back in my Windows partition. However, there has also been some help from Ivan Krylov on the r-package-devel email list:
I tried removing I looked at the Dependency Walker, but it was giving me many messages These libraries seem to be part of the UCRT system. However, when I looked at the DLL of a package that did load in Dependency Walker, I got the same messages, so I think the path to the UCRT files may be missing. Dependency Walker did seem to find: For NETIO.SYS it gave the " Error opening file. The system cannot find |
|
In a different post, Uwe Ligges suggested
I've queried:
Uwe's response:
Tomas Kalibera has responded to the same question:
It seems clear that if there is a need for the R Gtk bindings, a way has to be found of upgrading the package to work with at least Gtk3, and ultimately Gtk4. This would clearly be an investment of effort, so I'm going to check what the block with using the gWidgets2Tcl bindings was - the original reason for using RGtk2 (as well as it looking nicer). |
|
This old patch from Tomas Kalibera may be helpful: |
I'm trying to get RGtk to build on Windows, so that I can be confident that my package Retistruct will be usable in the future for Windows users (see this issue davidcsterratt/retistruct#52).
I was wondering if it would be possible to get RGtk to build using only the libraries available in Rtools, which appear to include all the required dependencies. It would then be possible (I suppose) to get new versions compiled using the win_builder service, which might assist with getting RGtk back on CRAN.
I've been experimenting with setting up the PKG_CPPLFAGS and PKG_LIBS in Makevars.win using the output from the pkg-config that I installed into the Rtools mingw32 system using pacman:
The source files compile to
.ofiles OK, but then there are linking issues - see truncated output below and the full output (attached). I fear I may be on a fool's errand, but any advice on what do to next would be much appreciated... I am reasonably experienced in building packages on Linux, but not on Windows.linking-error.txt
The text was updated successfully, but these errors were encountered: