5.0-beta13 with SAGE_CHECK=yes, Ubuntu 11.10 and 12.04
Dan Drake
Ubuntu. I figured out the environment variables to make the gcc spkg
happy with the multiarch stuff in Ubuntu. Since then I've observed that
builds with SAGE_CHECK=yes seem to fail. I've seen the ppl spkg hang,
taking up huge amounts of memory, and in other situations I've seen
building the documentation fail, by running the load average way, way
up.
Builds without SAGE_CHECK seem to finish.
Has anyone else seen such problems? For some of this, I also had
Virtualbox running, and it seems like Ubuntu 11.10 does not work so well
with Ubuntu 12.04 guests. I'm testing to see if that was the cause of
the problems, but I thought I would ask.
Dan
--
--- Dan Drake
----- http://mathsci.kaist.ac.kr/~drake
-------
Jeroen Demeyer
> I am encountering lots of problems with beta13 on recent releases of
> Ubuntu. I figured out the environment variables to make the gcc spkg
> happy with the multiarch stuff in Ubuntu.
> Since then I've observed that
> builds with SAGE_CHECK=yes seem to fail. I've seen the ppl spkg hang
I've seen some PPL test failures, yes.
See
http://groups.google.com/group/sage-release/t/9e0c3f12205a6344
Jeroen Demeyer
> building the documentation fail, by running the load average way, way
> up.
the docbuilding as far as I know.
Dan Drake
> On 2012-04-12 02:55, Dan Drake wrote:
> > I am encountering lots of problems with beta13 on recent releases of
> > Ubuntu. I figured out the environment variables to make the gcc spkg
> > happy with the multiarch stuff in Ubuntu.
> If you did, you should probably tell us :-)
I did: https://groups.google.com/d/msg/sage-release/FXYf_5g0B-4/C65HKR4EkLcJ
Jeroen Demeyer
> On Thu, 12 Apr 2012 at 08:30AM +0200, Jeroen Demeyer wrote:
>> On 2012-04-12 02:55, Dan Drake wrote:
>>> I am encountering lots of problems with beta13 on recent releases of
>>> Ubuntu. I figured out the environment variables to make the gcc spkg
>>> happy with the multiarch stuff in Ubuntu.
>> If you did, you should probably tell us :-)
>
> I did: https://groups.google.com/d/msg/sage-release/FXYf_5g0B-4/C65HKR4EkLcJ
As far as I can see, you never mentioned that you actually managed to
build all of Sage. So, you did?
Volker Braun
Dan Drake
Yes. Without SAGE_CHECK, I built beta13 on Ubuntu 11.10 (64-bit Core2
quad system, and 32-bit Atom netbook) and 12.04 (32-bit virtual machine
on my Core2 quad).
With SAGE_CHECK=yes, on the 11.10 64-bit system, the mipproblem stuff in
ppl sucks up 3+ gigabytes of memory. (It was an "as" process, as I
reported before.) Here is the last couple bits of the ppl log before I
killed it:
libtool: link: g++ -g -O2 -frounding-math
-I/scratch/sage-5.0.beta13/local/include -W -Wall -o .libs/mipproblem3
mipproblem3.o -L/scratch/sage-5.0.beta13/local/lib
../../utils/libppl_utils.a ../../tests/libppl_tests.a
../../src/.libs/libppl.so /scratch/sage-5.0.beta13/local/lib/libgmpxx.so
/scratch/sage-5.0.beta13/local/lib/../lib64/libstdc++.so -lm
/scratch/sage-5.0.beta13/local/lib/libgmp.so -Wl,-rpath
-Wl,/scratch/sage-5.0.beta13/local/lib -Wl,-rpath
-Wl,/scratch/sage-5.0.beta13/local/lib/../lib64 /bin/bash ../../libtool
--tag=CXX --mode=link g++ -g -O2 -frounding-math
-I/scratch/sage-5.0.beta13/local/include -W -Wall
-L/scratch/sage-5.0.beta13/local/lib -o mipproblem2
mipproblem2-mipproblem2.o ../../utils/libppl_utils.a
../../tests/libppl_tests.a ../../src/libppl.la
-L/scratch/sage-5.0.beta13/local/lib -lgmpxx
-L/scratch/sage-5.0.beta13/local/lib -lgmp
-R/scratch/sage-5.0.beta13/local/lib
-R/scratch/sage-5.0.beta13/local/lib ../../Watchdog/src/libpwl.la
libtool: link: g++ -g -O2 -frounding-math
-I/scratch/sage-5.0.beta13/local/include -W -Wall -o .libs/mipproblem2
mipproblem2-mipproblem2.o -L/scratch/sage-5.0.beta13/local/lib
../../utils/libppl_utils.a ../../tests/libppl_tests.a
../../src/.libs/libppl.so /scratch/sage-5.0.beta13/local/lib/libgmpxx.so
/scratch/sage-5.0.beta13/local/lib/libgmp.so
../../Watchdog/src/.libs/libpwl.so
/scratch/sage-5.0.beta13/local/lib/../lib64/libstdc++.so -lm -Wl,-rpath
-Wl,/scratch/sage-5.0.beta13/local/lib -Wl,-rpath
-Wl,/scratch/sage-5.0.beta13/local/lib/../lib64
With SAGE_CHECK=yes and SAGE_INSTALL_GCC=yes, everything builds and
passes doctests on my Ubuntu 10.04 Xeon machine with 8 gigabytes of
memory.
It does look like running the ppl tests will be problematic for Ubuntu
users. We also need to sort out the gcc spkg stuff for recent Ubuntu
releases; either set env vars correctly, or don't build gcc.
Dan Drake
I have not seen that either. I haven't been able to reproduce it,
though. I suspect it may have been a bad interaction with Virtualbox;
when I have VMs running, the rest of my system really slows down. Even
that sounds like it shouldn't affect building the documentation, though,
so I'll respond when I can reproduce.
leif
> Doc building is usually the least CPU intensive part since it is
> effectively single-threaded and done after everything else is finished.
> Which processes are causing the high load?
Well, you don't need CPU-intensive tasks to get a high sysload, as it
doesn't reflect the number of running, but /runnable/ processes.
So even if you have just a single process eating up all memory, the
sysload will naturally rise, as only occasionally running processes will
have their pages swapped out, hence cannot be run immediately, but count
as runnable.
With 'top' you'd in that case usually see a very low CPU usage
(user/system), while the load average in contrast can be quite high.
-leif
> On Thursday, April 12, 2012 7:33:08 AM UTC+1, Jeroen Demeyer wrote:
>
> On 2012-04-12 02:55, Dan Drake wrote:
> > I've seen
> > building the documentation fail, by running the load average way, way
> > up.
> This is something I have never seen. Also, SAGE_CHECK doesn't influence
> the docbuilding as far as I know.
>
> --
> You received this message because you are subscribed to the Google
> Groups "sage-release" group.
> To view this discussion on the web visit
> https://groups.google.com/d/msg/sage-release/-/t_XJfri412kJ.
> To post to this group, send email to sage-r...@googlegroups.com.
> To unsubscribe from this group, send email to
> sage-release...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/sage-release?hl=en.
--
() The ASCII Ribbon Campaign
/\ Help Cure HTML E-Mail
Dan Drake
> Volker Braun wrote:
> >Doc building is usually the least CPU intensive part since it is
> >effectively single-threaded and done after everything else is finished.
> >Which processes are causing the high load?
>
> Well, you don't need CPU-intensive tasks to get a high sysload, as
> it doesn't reflect the number of running, but /runnable/ processes.
yes. In my experience, when something goes crazy and my load average
spikes, it's because of disk access. Actual CPU usage is low, since (it
appears) a bunch of processes are fighting over the disk.
(What's strange is that I'm running builds with "nice -n19 ionice -c3
...", which seems like it would prevent such things, but apparently
not...)
Dan Drake
> With SAGE_CHECK=yes, on the 11.10 64-bit system, the mipproblem stuff in
> ppl sucks up 3+ gigabytes of memory.
Now I've tested this on some 32-bit systems, both using beta13's gcc spkg.
On my netbook (Ubuntu 11.10, 2 gigs RAM, Atom processor), the ppl spkg
built and passed tests.
On my Ubuntu 12.04 virtual machine (1.5 gigs RAM, Core2 processor), the
ppl spkg fails on one of the now-infamous mipproblem tests. Log is
attached.
Adam Webb
> Builds without SAGE_CHECK seem to finish.
>
> Has anyone else seen such problems? For some of this, I also had
> Virtualbox running, and it seems like Ubuntu 11.10 does not work so well
> with Ubuntu 12.04 guests. I'm testing to see if that was the cause of
> the problems, but I thought I would ask.
>
> Dan
>
> --
I had a problem with gcc but once I bypassed that (using touch) then
the build could continue. I also used SAGE_CHECK="yes". The build
finished and a 'make testlong' finished with no errors. I have not yet
tried on ubuntu 12.04. That is still changing on a daily basis
however.
Adam
Adam Webb
On Apr 12, 8:28 pm, Dan Drake <dr...@kaist.edu> wrote:
> On Thu, 12 Apr 2012 at 09:31PM +0900, Dan Drake wrote:
> > With SAGE_CHECK=yes, on the 11.10 64-bit system, the mipproblem stuff in
> > ppl sucks up 3+ gigabytes of memory.
>
> Now I've tested this on some 32-bit systems, both using beta13's gcc spkg.
>
> On my netbook (Ubuntu 11.10, 2 gigs RAM, Atom processor), the ppl spkg
> built and passed tests.
>
> On my Ubuntu 12.04 virtual machine (1.5 gigs RAM, Core2 processor), the
> ppl spkg fails on one of the now-infamous mipproblem tests. Log is
> attached.
>
> Dan
>
ppl. The log is not useful as it only says that the test failed. I
didn't note if there was a lot of memory used but the test did not
hang the VM or anything like that. I should mention that the VM only
has 2GB available to it so it might have just given up.
make check-TESTS
make[5]: Entering directory `/home/math/sage/spkg/build/ppl-0.11.2.p1/
src/tests/MIP_Problem'
PASS: ascii_dump_load1
PASS: exceptions1
PASS: mipproblem1
tests failed: test08
FAIL: mipproblem3
This is in wmware on Windows 7 (64 bit). I am running Ubuntu 12.04 (32
bit) which I updated yesterday.
Adam