Re: toolchain problems
Hi,
Is somebody taking on the ilmbase binNMU?
Done, and uploaded.
I'm thinking about building mesa with gcc-4.2 once we get below 150
needs-build. If that succeeds, it'll add several hundred to needs-build.
Any thoughts?
gcc
* courier_0.59.0-1 error: insn does not satisfy its constraints
* gforth_0.6.2-7.3 error: unable to find a register to spill in class
Reduce optimization?
* gmfsk_0.6+0.7pre1-2.1 internal compiler error: Segmentation fault
* llvm_2.2-7: internal compiler error: in find_reloads, at reload.c:3744
* openvrml_0.15.10-9.1 g++: Internal error: Killed (program cc1plus)
* osptoolkit_3.4.2-1 internal compiler error: in
reload_cse_simplify_operands, at postreload.c:395
That rings a bell - we had a gcc error in there before, maybe it got
reintroduced. I'll see what I can can find in my mailbox.
* pie: we've got a number of packages that fail to build because -pie
gives a nonrepresentational output. Can that option be a no-op or
something on m68k so at least stuff builds without m68k-specific
workarounds?
We can try that - hard to get that sneaked by the gcc maintainers though.
java: something is causing grief building java, I'm hoping it's ecj, which
is currently building, see charva, csound, icepick, jmagick,
libmatthew-java, sacjava, swt-gtk.
fortran: blas, self-test fails, see 478326
python2.5: hangs forever on test_bsddb3.
ghc6: virtual memory exhausted: Cannot allocate memory (I just requeued
this on a buildd with 768 MB RAM and an 800MB swap.)
May just be unbuildable with our hardware.
ruby1.9: semop test error, plus timeout
The timeout may be due to the semop error ...
Looking at blas made me realize that I've somehow totally forgotten
fortran. Maybe those were weak braincells and they died.
Nah, these were the _strong_ brain cells than managed to forget :-) I'll
see how much fortran I can muster.
I've got another 48 or so failures to classify.
Don't give up just yet :-)
Michael
Reply to: