Bug 1164209 - hevea segfaults when creating docs for ejabberd
Summary: hevea segfaults when creating docs for ejabberd
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: ocaml
Version: 21
Hardware: ppc64
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: Richard W.M. Jones
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Keywords:
Depends On:
Blocks: PPCTracker F-ExcludeArch-ppc64le, PPC64LETracker
TreeView+ depends on / blocked
 
Reported: 2014-11-14 11:05 UTC by Dan Horák
Modified: 2014-11-18 12:31 UTC (History)
3 users (show)

(edit)
Clone Of:
(edit)
Last Closed: 2014-11-18 12:31:31 UTC


Attachments (Terms of Use)

Description Dan Horák 2014-11-14 11:05:14 UTC
hevea segfaults when creating docs for ejabberd, see eg. http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=2181864 or http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=2181914


Version-Release number of selected component (if applicable):
hevea-2.16-2.fc21.ppc64le
hevea-2.16-2.fc21.ppc64

Comment 1 Richard W.M. Jones 2014-11-14 11:11:08 UTC
Do you happen to know if this happens on F22 as well?

CC-ing Michel Normand who wrote the ppc64le backend.

Comment 2 Richard W.M. Jones 2014-11-14 11:12:53 UTC
The other thing to try would be increasing the stack size, since
ppc64 code often needs a lot more stack than is available by default.

Put something like:

%ifarch ppc64 ppc64le
ulimit -a
ulimit -Hs 65536
ulimit -Ss 65536
%endif

(or larger even) before the make.

Comment 3 Dan Horák 2014-11-14 11:38:41 UTC
yeah, the larger stack helps - http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=2182090

diff --git a/ejabberd.spec b/ejabberd.spec
index 14d374a..55a9e26 100644
--- a/ejabberd.spec
+++ b/ejabberd.spec
@@ -150,6 +150,11 @@ autoreconf -ivf
 make
 %if 0%{?_with_hevea}
 pushd doc
+%ifarch %{power64}
+ulimit -a
+ulimit -Hs 65536
+ulimit -Ss 65536
+%endif
 # remove pre-built docs
 rm -f dev.html features.html
 # See this link - http://thread.gmane.org/gmane.linux.redhat.fedora.devel/198954/focus=198957

Comment 4 Richard W.M. Jones 2014-11-14 11:47:29 UTC
I believe you can coalesce the two ulimit statements into just:

  ulimit -s 65536

This is not the first package we've had to do this for.  I
wonder if we can increase the default stack size on ppc64/le?
Honestly I've never seen the point of limiting stack in the
first place - we don't limit any other type of memory by default.

Comment 5 Dan Horák 2014-11-14 11:57:34 UTC
I have already seen also on s390(x) that the default stack was too small. Is the limit a kernel default? I think there is a relation between the default stack size and the number of available processes/thread before they get memory exhausted. But I agree maybe it is the right time to change the too low deaults.

Comment 6 Fedora Update System 2014-11-14 12:14:06 UTC
ejabberd-14.07-6.fc21 has been submitted as an update for Fedora 21.
https://admin.fedoraproject.org/updates/ejabberd-14.07-6.fc21

Comment 7 Richard W.M. Jones 2014-11-14 12:38:58 UTC
I'm guessing it is a kernel default.  However the best place
to change the default might be in /etc/security/limits.conf
which is in the 'pam' package.  I'm not sure if limits.conf
applies only to logins, or to every process.

Comment 8 Fedora Update System 2014-11-15 09:22:24 UTC
Package ejabberd-14.07-6.fc21:
* should fix your issue,
* was pushed to the Fedora 21 testing repository,
* should be available at your local mirror within two days.
Update it with:
# su -c 'yum update --enablerepo=updates-testing ejabberd-14.07-6.fc21'
as soon as you are able to.
Please go to the following url:
https://admin.fedoraproject.org/updates/FEDORA-2014-15140/ejabberd-14.07-6.fc21
then log in and leave karma (feedback).

Comment 9 Fedora Update System 2014-11-18 12:31:31 UTC
ejabberd-14.07-6.fc21 has been pushed to the Fedora 21 stable repository.  If problems still persist, please make note of it in this bug report.


Note You need to log in before you can comment on or make changes to this bug.