I tried to build R on AIX and after 2 minor bugfixes everything seems to work fine except for dynamic loading of code. In a previous mail I found that Fabrizio Moro succeeded using SHLIBLDFLAGS="-bM:S1L -bnoentry -bnogc -berok -lc -lm -lodm -lcfg" which compiles fine for me, I get also no error messages upon dynloading, but no symbols are available afterwards, i.e. is.loaded fails (I tried also with trailing underscores etc.). Has anybody some experience which flags to use? I use an rs6000-ibm-aix4.2.1.0 and the native cc, f77 and ld (gcc compiles but produces incorrect code, i.e., some computations are wrong). Make check succeeds until eda :-) .f -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>>>>> "FrL" == Friedrich Leisch <Friedrich.Leisch@ci.tuwien.ac.at> writes:FrL> I tried to build R on AIX and after 2 minor bugfixes everything FrL> seems to work fine except for dynamic loading of code. FrL> In a previous mail I found that Fabrizio Moro succeeded using FrL> SHLIBLDFLAGS="-bM:S1L -bnoentry -bnogc -berok -lc -lm -lodm -lcfg" FrL> which compiles fine for me, I get also no error messages upon FrL> dynloading, but no symbols are available afterwards, FrL> i.e. is.loaded fails (I tried also with trailing underscores FrL> etc.). Yes. Unfortunately, Fabrizio (nor me) followed up on that mail telling you that we (Fabrizio and me) ended up with the exact same symptomps as you have. I gave up at that time; Fabrizio continues to use S-plus on AIX (and R on Linux, to some extentes at least)... ... yes a sad story, but I think we both ended up with the impression that AIX is half broken environment anyway --- at least if you want to compile software that has no problems at all on other architectures... FrL> Has anybody some experience which flags to use? FrL> I use an rs6000-ibm-aix4.2.1.0 and the native cc, f77 and ld (gcc FrL> compiles but produces incorrect code, i.e., some computations are FrL> wrong). FrL> Make check succeeds until eda :-) exactly. The first time, dyn.load is needed.... Martin. -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>>>>> Martin Maechler writes:>>>>> "FrL" == Friedrich Leisch <Friedrich.Leisch@ci.tuwien.ac.at> writes:FrL> I tried to build R on AIX and after 2 minor bugfixes everything FrL> seems to work fine except for dynamic loading of code. FrL> In a previous mail I found that Fabrizio Moro succeeded using FrL> SHLIBLDFLAGS="-bM:S1L -bnoentry -bnogc -berok -lc -lm -lodm -lcfg" FrL> which compiles fine for me, I get also no error messages upon FrL> dynloading, but no symbols are available afterwards, FrL> i.e. is.loaded fails (I tried also with trailing underscores FrL> etc.).> Yes. Unfortunately, Fabrizio (nor me) followed up on that mail telling you > that we (Fabrizio and me) ended up with the exact same symptomps as you > have.> I gave up at that time; Fabrizio continues to use S-plus on AIX > (and R on Linux, to some extentes at least)... > ... yes a sad story, but I think we both ended up with the impression that > AIX is half broken environment anyway > --- at least if you want to compile software that has no problems at all > on other architectures...FrL> Has anybody some experience which flags to use? FrL> I use an rs6000-ibm-aix4.2.1.0 and the native cc, f77 and ld (gcc FrL> compiles but produces incorrect code, i.e., some computations are FrL> wrong). FrL> Make check succeeds until eda :-)> exactly. The first time, dyn.load is needed....It might help to look into other software for which dynloading works under AIX. Three potential examples come to mind: Xlisp-stat Octave Tcl Anyone willing to have a look? -k -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
We have R running on AIX (several OS versions between 4.1 and 4.3) and we
are able to load code dynamically. Our solution was essentially developed
by Heiner Schwarte. We use a different version of dynload.c and the linker
is called like this:
$(F77) -o module.so \
-bM:SRE -bE:module.exp -bnoentry $(OBJECTS) $(LIBDIR) $(LIBS)
where module.exp contains all symbols that have to be exported.
Arne
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._