On Tue, Apr 17, 2007 at 10:07:55PM +0200, Luk Claes wrote:
> Dirk Eddelbuettel wrote:
> > Hi Luk,
> >
> > On Tue, Apr 17, 2007 at 07:20:08PM +0200, Luk Claes wrote:
> >> Package: quantlib-swig
> >> Severity: serious
> >> Version: 0.3.14-1
> >>
> >> Hi
> >>
> >> Your package fails to build from source on mips and mipsel due to
> >> relocation
> >> errors according to the build logs. Note that this indicates that some
> >> part of
> >> the code (or library build dependencies) might not be compiled with the PIC
> >> (Position Independent Code) flag enabled...
> >
> > Could that not also be indicative of a bug in the toolchain? After all,
> >
> > i) it used to build on these arches with the same flags, build-depends, ...
>
> New upstream versions... you might want to test compiling 0.3.13 with current
> build dependencies?
>
> > ii) it builds on the other arches
>
> relocation errors are more likely to occur on mips than on other
> architectures, because it's more strict in this regard.
>
> > What do you think?
>
> I think a toolchain bug is unlikely, though maybe it's caused by quantlib or
> boost.
Not QuantLib. Remember that this is a swig package -- there is exactly
_one_ call to g++ in the build log (indented for mail)
g++ -fno-strict-aliasing -DNDEBUG -g -O2 -Wall -Wstrict-prototypes -O0 \
-g0 -fPIC -I/usr/include -I/usr/include/python2.4 -c \
QuantLib/quantlib_wrap.cpp -o \
build/temp.linux-mips-2.4/QuantLib/quantlib_wrap.o -Wno-unused -O0 -g0
and given that this comes from the QuantLib / Swig / Python templates,
it also hasn't changed in a while. And of course contains -fPIC.
It is a monster compilation though: this slows i86 down as well and
consumes tons of memory due to the heavy templating.
I still think it is a toolchain issue.
Dirk
--
Hell, there are no rules here - we're trying to accomplish something.
-- Thomas A. Edison
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
