Summary: | dev-lang/R segfaults during compilation in modules/lapack | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Markus Luisser <mluisser> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | RESOLVED INVALID | ||
Severity: | blocker | ||
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Markus Luisser
2004-03-06 08:45:11 UTC
Can you reproduce exactly this error every time you try, or does it happen in various code lines? Could you maybe retry with more conservative optimizations? A third thought: If you want a math related package, do you really think -ffast-math is a good idea? This can result in incorrect calculations. yep - the error is reproducible at this line - i also tried to go down with the flags: ----------------snip---------------- # CFLAGS="-march=athlon -O2" CXXFLAGS="-march=athlon -O2" emerge R [...lots of output...] gcc -I. -I../../../src/include -I../../../src/include -I/usr/local/include -DHAVE_CONFIG_H -D__NO_MATH_INLINES -mieee-fp -fPIC -march=athlon -O2 -c sockconn.c -o sockconn.lo gcc -shared -L/usr/local/lib -o internet.so Rsock.lo internet.lo nanoftp.lo nanohttp.lo sock.lo sockconn.lo -L../../../bin -lR make[5]: Entering directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/internet' make[5]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/internet' make[4]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/internet' make[3]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/internet' make[3]: Entering directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/lapack' make[4]: Entering directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/lapack' making Lapack.d from Lapack.c make[4]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/lapack' make[4]: Entering directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/lapack' g77 -mieee-fp -fPIC -g -O2 -c dlapack0.f -o dlapack0.lo dlapack0.f: In subroutine `dlag2': dlapack0.f:3014: internal compiler error: Segmentation fault Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://bugs.gentoo.org/> for instructions. make[4]: *** [dlapack0.lo] Error 1 make[4]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/lapack' make[3]: *** [R] Error 2 make[3]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules/lapack' make[2]: *** [R] Error 1 make[2]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src/modules' make[1]: *** [R] Error 1 make[1]: Leaving directory `/var/tmp/portage/R-1.8.0/work/R-1.8.0/src' make: *** [R] Error 1 !!! ERROR: dev-lang/R-1.8.0 failed. !!! Function src_compile, Line 68, Exitcode 2 !!! (no error message) ----------------snip---------------- But i didnt expect something other than this as gcc fails with 'g77' and for the fortran code the flags are differently anyway. The thing about '-ffast-math' is a point - never thought much about it but as most of my calculations must hold true against measured values, i can live with the risk. So far i did not have any problems and i need every little bit of speed on this box :( I'll reconsider however if ffastmath has influenced my system so much that i cannot emerge 'R' (perhaps glibc i borked?) - what do you propose? I'd like to know the settings of your USE flags first... what's the result of emerge -pv R on your system? # emerge -pv R These are the packages that I would merge, in order: Calculating dependencies ...done! [ebuild N ] dev-lang/R-1.8.0 +X -atlas -gnome +tcltk 0 kB i meanwhile emerged R on another box without the --ffast-math flag and had no problems. Unfortunately i cannot give you the configuration of the box now because shortly after i emerged my ISP let me down. When i remember correctly the USE flags and the rest of the CFLAGS are quite similar to this box. When i have access to the net again, i'll post the config of the working box if this helps you in any way. i'm getting segmentation faults for other packages too now - i more and more tend to think that something is wrong with the box - unless someone else reports this, please do not invest work here - i'll try to sort this out and keep you updated - may take a while though, compiling is slow here... Ok - yesterday night i managed to recover from a major filesystem crash just one day after the power supply died. As nobody else seems to have issues with this i blame the flaky hardware and close it. |