On 2018-09-09 11:56 AM, Dirk Eddelbuettel wrote:
Thanks for the hint and its testing!
I'll add this. Before I do so can you quickly check if the tests in
debian/rules identify the platform correctly? Ie did you test it with this
block:
## edd 01 Feb 2002 allow -ffunction-sections on hppa
## edd 26 Oct 2007 turn strict aliasing warning off
## edd 01 May 2011 add -fpermissive as g++ is more of a stickler
## edd 06 May 2011 remove -fpermissive with QL 1.1 betas
compilerflags = -O2 -Wall -Wno-strict-aliasing -DBOOST_NO_AUTO_PTR
ifeq ($(cpu),hppa)
compilerflags = -O2 -ffunction-sections -DBOOST_NO_AUTO_PTR
#cxxcompiler = g++-3.0
endif
No. I modified this hunk which comes later:
## edd 22 Aug 2018 trying to help a few other systems with their
memory exhaustion
ifneq "$(findstring $(cpu), hppa hurd-i386 kfreebsd-i386 m68k powerpc
x32)" ""
compilerflags = -O0 -g0 --param ggc-min-expand=20 -DBOOST_NO_AUTO_PTR
endif
It overrides the previous define for compilerflags on hppa. I think we
need to do a couple
more builds. If -ffunction-sections fixes the stub table problem, it's
a better solution than
-mlong-calls.
We also need to determine whether "-O0 -g0 --param ggc-min-expand=20" is
needed to avoid
memory exhaustion.
The build log for the uploaded build is here:
https://buildd.debian.org/status/fetch.php?pkg=quantlib-swig&arch=hppa&ver=1.13-5&stamp=1536453367&raw=0
You can see the flags used:
#(cd Ruby && \
# CXX="g++" \
# CFLAGS="-O0 -g0 --param ggc-min-expand=20 -DBOOST_NO_AUTO_PTR
-mlong-calls" \
# CXXFLAGS="-O0 -g0 --param ggc-min-expand=20 -DBOOST_NO_AUTO_PTR
-mlong-calls" \
Dave
--
John David Anglin dave.ang...@bell.net
