make flintqs-20070817 spkg build with -m64 rather than -march=opteron

[sagetrac-mariah]

sage-4.7.rc2 built on an opteron system [skynet:flavius (x86_64-Linux-k8-fc)] with SAGE_FAT_BINARY="yes", and then moved to another x86_64-Linux system will fail with a segmentation fault on the following code:

set_random_seed(0)
p = next_prime(randrange(2^96))
q = next_prime(randrange(2^97))
n = p * q
qsieve(n) 

qsieve is from the flintqs-20070817 spkg, which by default builds
with -march=opteron on all x86_64 systems (and in 32-bit mode).

---

New spkg: http://sage.math.washington.edu/home/leif/Sage/spkgs/flintqs-20070817.p6.spkg

Component: packages: standard

Keywords: sd32 quadratic sieve

Author: Mariah Lenox

Reviewer: Martin Albrecht

Merged: sage-4.7.2.alpha3

Issue created by migration from https://trac.sagemath.org/ticket/11351

[sagetrac-mariah]

Attachment: flintqs-20070817.p6.spkg.gz

[sagetrac-mariah]

diff for review purposes only

[sagetrac-mariah]

Author: Mariah Lenox

[sagetrac-mariah]

Description changed:

--- 
+++ 
@@ -12,3 +12,5 @@
 
 'qsieve' is from the flintqs-20070817 spkg, which by default builds
 with -march=opteron on *all* x86_64 systems (*and* in 32-bit mode).
+
+**Apply** [attachment: flintqs-20070817.p6.spkg]

[sagetrac-mariah]

comment:1

Attachment: flintqs-20070817.p5-p6.diff.gz

[nexttime]

comment:2

Quite messy.

We should just apply a single patch to the Makefile, and set / pass the appropriate variables (CXXFLAGS, CXXINCS -- usually CPPFLAGS, LIBS, perhaps CXXFLAG64, which we can put into CXXFLAGS).

Also, $(CPP) should be $(CXX); the FLINT (1.5.x) Makefile is btw. equally bad.

-m64 or an analogue (if SAGE64=yes) is only necessary on 64-bit platforms that default to 32-bit builds, i.e. SunOS / Solaris and MacOS X 10.5.

Using stdint.h, uint32_t and uint64_t would be more portable, too.

[koffie]

comment:3

dear leif,

Your comment sounds like there is a lot still to be done. Do you think this tickets needs to change it's status to "needs_work" or do you want to give it a positive review and open a new ticket for your comment?

[malb]

Reviewer: Martin Albrecht

[malb]

comment:4

Looks good to me.