Debugging pexpect issues with Maxima/clisp/....
mabshoff
in the past couple moths we have seen many reports that indicate that
we are still having pexpect hangs, especially with Maxima. Since none
of those happen on systems where we usually have direct access it is
very difficult the debug the problem. A while ago we introduced some
settings that will make debugging these somewhat easier since the
traffic of such failures is logged on demand.
So if you encounter timeouts and they are clearly not caused by a slow
system please do the following.
First
export SAGE_PEXPECT_LOG=yes
this environment variable will cause all pexpect traffic to be logged
to files in $DOT_SAGE/pexpect_log, i.e. usually $HOME/.sage/
pexpect_log. If you don't use bash you have to use setenv in case you
use [t]csh/ksh, but I would assume there is something wrong with you
for using such a shell in the first place :p
Then identify all tests that give you timeouts during a "make check".
const.tex is a common one it seems and so is calculus.py. For example
now run
./sage -t -verbose devel/doc/const/const.tex
<SNIP>
At some point this should hang while telling you which test is
running. Now look in $DOT_SAGE/pexpect_log. You should see something
like the following:
singular-31329-14103328-2.log
gap-31329-14105248-2.log
pari-31329-73925648-2.log
pari-31329-73925168-2.log
maxima-31329-27545184-2.log
maxima-31329-15330912-2.log
Identify the youngest file (ls -latr) and upload it to some place
where it can be downloaded. Since those files tend to be large do not
attach them to tickets. Then copy and paste the last 20 or so lines
from the doctest output into a sage-devel thread with a meaning ful
summary, give the system info and anything else that might be useful
and post a link to the log in that message. One of us will take a
closer look and once we have identified the problem we will open a
ticket and work on a fix.
I will not put this info into the wiki since it belongs in the
documentation, so I will add it post Sage 3.4. Since we are also
switching to ecl in Sage 3.4 I expect various other pexpect related
problem to crop up in and post 3.4, so the above should come in
handy.
Cheers,
Michael
Burcin Erocal
On Fri, 13 Feb 2009 06:26:53 -0800 (PST)
mabshoff <mabs...@googlemail.com> wrote:
<snip>
> So if you encounter timeouts and they are clearly not caused by a slow
> system please do the following.
>
> First
>
> export SAGE_PEXPECT_LOG=yes
I attached a log from a test instance that timed out with 3.4. I don't
see any obvious problems, but I have no experience with the pexpect
interfaces.
There is a longer log, which contains an identical problem in the
beginning, and record of me hitting CTRL+C, then the rest of the tests
running as expected here:
http://sage.math.washington.edu/home/burcin/maxima-18304-31961408-2.log
Let me know what else I can do to debug this. The maxima pexpect
interface never worked on this laptop and it's really annoying trying
to develop symbolics without running doctests.
Cheers,
Burcin