[Rd] MinGW and the ld bug

[Prof Brian Ripley]

I believe the bug is triggered by the fairly unusual event you have a '.' 
in the package name, and the change in 2.2.1 to 2.2.1 patched was to add

> LIBRARY R.native.dll
                   ^^^^
(which is what my workaround in ld.exe was doing another way).

Since we are told that is the correct form and we were using it 
incorrectly before, I don't really know how to work around this.  For you, 
just delete .dll in the following line in src/gnuwin32/MkRules

%.dll:
 	@$(ECHO) LIBRARY $*.dll > $*.def

but that is not a general solution.


On Mon, 13 Feb 2006, Henrik Bengtsson wrote:

> Hi.
>
> I noticed that Brian Ripley found and corrected a bug in MinGW's
> ld.exe, see http://www.murdoch-sutherland.com/Rtools/.  Thanks for
> this.  I wonder if this is the same bug that cause my problems.  I
> have tiny toy package with C code that installs perfectly on R Version
> 2.2.1 beta (2005-12-18 r36792) [this version was mislabelled "beta"
> the first few hours on CRAN when the stable 2.2.1 version first came
> out].  However, when I try to install the same on R Version 2.2.1
> Patched (2006-02-09 r37305) as well as R Version 2.3.0 Under
> development (unstable) (2006-02-02 r37243), I get the following error:
>
> C:\Documents and Settings\hb\braju.com.R\R.native\build>Rcmd install R.native
>
> ---------- Making package R.native ------------
>  adding build stamp to DESCRIPTION
>  making DLL ...
> making rowMedians.d from rowMedians.c
> gcc   -IC:/PROGRA~1/R/R-2.2.1pat/include -Wall -O2   -c rowMedians.c -o rowMedia
> ns.o
> ar cr R.native.a rowMedians.o
> ranlib R.native.a
> windres --include-dir C:/PROGRA~1/R/R-2.2.1pat/include  -i R.native_res.rc -o R.
> native_res.o
> gcc  --shared -s  -o R.native.dll R.native.def R.native.a R.native_res.o  -LC:/P
> ROGRA~1/R/R-2.2.1pat/src/gnuwin32   -lg2c -lR
> c:\MinGW\bin\..\lib\gcc\mingw32\3.4.4\..\..\..\..\mingw32\bin\ld.exe: R.native.d
> ef:1: syntax error
> c:\MinGW\bin\..\lib\gcc\mingw32\3.4.4\..\..\..\..\mingw32\bin\ld.exe:R.native.de
> f: file format not recognized; treating as linker script
> c:\MinGW\bin\..\lib\gcc\mingw32\3.4.4\..\..\..\..\mingw32\bin\ld.exe:R.native.de
> f:1: syntax error
> collect2: ld returned 1 exit status
> make[3]: *** [R.native.dll] Error 1
> make[2]: *** [srcDynlib] Error 2
> make[1]: *** [all] Error 2
> make: *** [pkg-R.native] Error 2
> *** Installation of R.native failed ***
>
> Removing 'C:/PROGRA~1/R/R-2.2.1pat/library/R.native'
> Restoring previous 'C:/PROGRA~1/R/R-2.2.1pat/library/R.native'
>
> My R.native.def looks like this:
>
> LIBRARY R.native.dll
> EXPORTS
> rowMedians
> rowMediansInteger
> rowMediansReal
>
> I've tried to replace the two occurances of ld.exe in MINGW with BR's
> patch, but I get the same error.  The ld of the patch and the original
> one show the same version string;
> patch: GNU ld version 2.16.91 20050827 and original: GNU ld version
> 2.16.91 20050827. I've tried to use both MINGW v5.0.0 and v5.0.2 with
> both "current" (v3.4.2) and "candidate" (v3.4.4) packages.

Hmm, only 3.4.5 is there as a candidate.

> Is this related to the bug BR found?  Any suggestions what might go
> wrong between R v2.2.1 and v2.2.1 patched?
>
> Thanks
>
> Henrik
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595