More quantlib-python failures on non-i386

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

More quantlib-python failures on non-i386

Dirk Eddelbuettel
You might recall that Luigi debugged QL-Python on alpha.  Lots of other
architectures are still ailing, though.  As

shows, we currently fail on alpha, hppa, ia64, arm, m68k. As I read the build
logs, a quick summary is

alpha   wrong (0.2.1, not 0.2.1cvs) library installed, not attempted
hppa    needs -ffunction-sections compiler switch [ see below ]
ia64    some compiler warnings, survives few tests, then fails distributions
arm too small a machine, aborts after 150 mins thinking it failed
m68k    idem, too small / too little memory to cope with the complex C++

I should get the alpha to attempt a recompilation [ though we currently have
hardware problems with our main Alpha box ], and get the timeout increased
for arm and m68k. [1]  That would leave hppa and ia64.  

I actually modified the build script for hppa earlier to add the
"-ffunction-sections" switch for gcc. It now builds, but then fails a few
tests (details below in [2]).  This is based on on the 0.3.0a5 branch from
CVS which I called 0.2.1cvs20020120 [ Debian needs to sort the version nb ]

Luigi: James Troup, who acts as Debian sysadmin for this hppa box, could
create an account for you if you send him your ssh public key.  Could you do
that, and then try to squash the bug there?

Cheers,  Dirk

[1] Or simply prevent it from being built, but that is considered sacrilege
within Debian as everything ought to build everywhere, something I personally
disagree with.

[2] On hppa, the following tests all fail:

Whereas these work fine:

Good judgment comes from experience; experience comes from bad judgment.
                                                            -- F. Brooks