On 2010-Jul-28 00:02:48 +1200, François Bissey <[email protected]> wrote:
>Actually the comentator from linbox breaks sage. I thought this was
>known and patched in the linbox spkg. I just checked the linbox
>spkg and it is removed.

Can you tell me what you are checking here.

>That's strange that you seem to have built a linbox with commentator.

I have compared my build log with a 4.5.2alpha0 build on boxen and
there are no relevant differences.  In particular, commentator.o is
built and embedded into libutil.a (and hence liblinbox.so) on both
systems.  I have no local patches to linbox-1.1.6.p3.  The spkg has a
patch to commentator.C to patch out the default global Commentator -
which has been applied on both systems.

I have had a look at trac #2833 (which was the origin of the above
patch) and my linbox compile succeeds so the symptoms don't match
those in #2833.

On 2010-Jul-27 13:57:06 +0100, David Kirkby <[email protected]> wrote:
>On 27 July 2010 12:36, Peter Jeremy <[email protected]> wrote:
>> I've managed to get Sage 4.5 to build on FreeBSD (see
>> http://wiki.sagemath.org/freebsd/sage-4.5 for details)
>
>Well done. Having another platform is likely to show bugs not seen on
>any current platform.

I've reached this point in the past but couldn't keep forward-porting
my fixes to keep up with the rate at which Sage is released.

>> seems to run (I'm still waiting for the test suite results)

BTW, the test invocation in the main makefile still includes the
assumption that /bin/sh is bash - I therefore had to run it manually.
I hope to come up with a fix.

Apart from the segmentation violations, the test suite seems to
mostly pass - which is good.

>> but it consistently gets segmentation violations on exit.
>
>I get exactly the same with a 64-bit build on Solaris, though I don't
>think the issue is connected with what you see - my problem seems more
>Pari or Singular related.
>
>There's a comment here about the linbox
>
>http://groups.google.co.uk/group/sage-devel/browse_thread/thread/a5408a037adacd47/356c72a50b1e33e7?lnk=gst&q=solaris+64-bit#356c72a50b1e33e7
>
>that might be worth reading.

I don't see anything relevant there - linbox compiles with no errors
or warnings.

>I would build without optimisation to see if that helps.

Most of linbox is already built without optimisation (commentator.C
definitely is).

Out of interest, I tried building with a different gcc/g++ but this
ran into problems with linbox dependencies requiring gcc4.5.

> It wold be worth adding a spkg-check if
>there's not one.

Well, I've manually run the self-tests in 'make check' and, apart from
the one test that won't compile, all pass.

>Can you confirm that there is a make check in linbox?

Yes, there is, though test-gf2.C fails to compile with:
g++ -DHAVE_CONFIG_H -I. -I.. -I.. -I../linbox -D__LINBOX_HAVE_CBLAS 
-I/tank/obj/sage/sage-4.5/local/include -I/tank/obj/sage/sage-4.5/local/include 
-I/tank/obj/sage/sage-4.5/local/include  -g 
-I"/tank/obj/sage/sage-4.5/local/include/linbox" 
-I"/tank/obj/sage/sage-4.5/local"/include -g -DLinBoxSrcOnly -Wall  -g -fPIC 
-I"/tank/obj/sage/sage-4.5/local/include" 
-I"/tank/obj/sage/sage-4.5/local/include/linbox"  
-L"/tank/obj/sage/sage-4.5/local/lib" 
-I/tank/obj/sage/sage-4.5/spkg/build/linbox-1.1.6.p3/src 
-I/tank/obj/sage/sage-4.5/spkg/build/linbox-1.1.6.p3/src/linbox  
-I/tank/obj/sage/sage-4.5/local/include  
-I/tank/obj/sage/sage-4.5/local/include -D__LINBOX_HAVE_CBLAS -c -o test-gf2.o 
test-gf2.C
In file included from ../linbox/vector/bit-vector.h:170:0,
                 from ../linbox/field/gf2.h:23,
                 from test-gf2.C:26:
../linbox/vector/bit-vector.inl: In member function 
'LinBox::BitVector::const_iterator& 
LinBox::BitVector::const_iterator::operator=(const 
std::iterator<std::random_access_iterator_tag, bool>&)':
../linbox/vector/bit-vector.inl:258:18: error: 'const struct 
std::iterator<std::random_access_iterator_tag, bool>' has no member named '_ref'
../linbox/vector/bit-vector.inl:259:17: error: 'const struct 
std::iterator<std::random_access_iterator_tag, bool>' has no member named '_ref'
../linbox/vector/bit-vector.inl: In member function 'bool 
LinBox::BitVector::const_iterator::operator==(const 
std::iterator<std::random_access_iterator_tag, bool>&) const':
../linbox/vector/bit-vector.inl:336:29: error: 'const struct 
std::iterator<std::random_access_iterator_tag, bool>' has no member named '_ref'
../linbox/vector/bit-vector.inl:336:60: error: 'const struct 
std::iterator<std::random_access_iterator_tag, bool>' has no member named '_ref'
../linbox/vector/bit-vector.inl: In member function 'bool 
LinBox::BitVector::const_iterator::operator!=(const 
std::iterator<std::random_access_iterator_tag, bool>&) const':
../linbox/vector/bit-vector.inl:342:29: error: 'const struct 
std::iterator<std::random_access_iterator_tag, bool>' has no member named '_ref'
../linbox/vector/bit-vector.inl:342:60: error: 'const struct 
std::iterator<std::random_access_iterator_tag, bool>' has no member named '_ref'
../linbox/vector/bit-vector.inl:342:72: warning: control reaches end of 
non-void function
make[3]: *** [test-gf2.o] Error 1

The remainder of "make check" compiles and passes.

-- 
Peter Jeremy

Attachment: pgpXWXTmbbN87.pgp
Description: PGP signature

Reply via email to