[Bug-report]
[Lunar Linux 0000290]: Mozilla 1.3 fails (hangs) during post-install
bug-reports at lunar-linux.org
bug-reports at lunar-linux.org
Fri May 16 20:55:40 GMT 2003
A BUGNOTE has been added to this bug.
=======================================================================
http://bugs.lunar-linux.org/view_bug_page.php?f_id=0000290
=======================================================================
Reporter: cpowers
Handler: nhudson
=======================================================================
Project: Lunar Linux
Bug ID: 0000290
Category: moonbase
Reproducibility: always
Severity: minor
Priority: normal
Status: assigned
=======================================================================
Date Submitted: 2003-05-08 05:03 GMT
Last Modified: 2003-05-16 20:55 GMT
=======================================================================
Summary: Mozilla 1.3 fails (hangs) during post-install
Description:
Mozilla 1.3 sucessfully compiles, installs, and creates the
compile and install logs but never goes on to create the cache. It hangs
at the "Creating ...install/mozilla-1.3"
output line for as long as I'll let it (hours). When I finally kill it,
the following output is produced:
Creating /var/log/lunar/compile/mozilla-1.3.bz2
Creating /var/log/lunar/install/mozilla-1.3
//sbin/lin: line 960: 22505 Terminated (
CACHE="/tmp/$MODULE-$VERSION-$BUILD.tar";
CACHE_BZ="$INSTALL_CACHE/$MODULE-$VERSION-$BUILD.tar.bz2";
C_LOG_BZ="$COMPILE_LOGS/$MODULE-$VERSION.bz2";
TMP_LOG="/tmp/$MODULE-$VERSION";
INST_LOG="$INSTALL_LOGS/$MODULE-$VERSION";
MD5_LOG="$MD5SUM_LOGS/$MODULE-$VERSION"; IW_LOG="/tmp/$MODULE.iw"; rm -f
$CACHE; rm -f $INST_LOG; touch $INST_LOG; rm -f $TMP_LOG; rm -f $MD5_LOG;
touch $MD5_LOG; track && archive && add_module $MODULE
installed $VERSION `find_module_size $MODULE $VERSION`; rm -f $BOOST_LOCK
)
! Problem detected during "boost" !
=======================================================================
-----------------------------------------------------------------------
nhudson - 2003-05-09 22:18 GMT
-----------------------------------------------------------------------
Mozilla 1.3 is no longer in moonbase it has been updated to 1.3.1, I have
tested mozilla builds over and over again on 3 diffrent machines over the
last day or so and all worked with no hitches glitches. If you are haing
this problem I would say something is wrong here with your box, possibly
hardware related, bad memory, bad power supply. I would try the updated
module before posting again to see if it works
-----------------------------------------------------------------------
cpowers - 2003-05-11 01:54 GMT
-----------------------------------------------------------------------
Sorry to say but it looks like 1.3.1 is doing the exact same thing
as 1.3. I just finished linning 1.3.1, with '-c -r ' options, and
according to ps, I'm using it now. But, the install window is just
sitting there, doing nothing for the past half hour, after displaying
the 'Creating /var/log/lunar/install/mozilla-1.3.1' entry. Lunar has,
in fact, created all 3 logs, to wit:
-rw-r--r-- 1 root root 214825 May 10 18:44
compile/mozilla-1.3.1.bz2
-rw-r--r-- 1 root root 1222799 May 10 18:46
install/mozilla-1.3.1
-rw-r--r-- 1 root root 96021 May 10 18:46
md5sum/mozilla-1.3.1
but does not continue on to create the cache and write to the activity
log.
The only entry in the activity log will be a "failed during boos" entry
when
I get tired of seeing the process not doing anything, and kill it.
Here are the prior activity log enties for mozilla 1.3 on this computer.
Note that this problem showed up AFTER your recent Lunar bug fix release.
Did something in lin, line 960, change in or after your bug-fix release?
20030412-15:35:43 lin mozilla 1.3 success
20030503-22:18:12 lrm mozilla 1.3 success
20030504-02:59:46 lin mozilla 1.3 failed Due to build
errors during boost
20030504-12:21:48 lin mozilla 1.3 failed Due to build
errors during boost
The same is true on my second computer. I wonder if I could manually
edit
the activity log to make Lunar think the installation suceeded? I
wouldn't
have the cache, but could tar up the src directory. I moght try that on
my other computer come Monday.
Sorry to be such a pest. I like Lunar and I appreciate your taking time
to
look into this.
-----------------------------------------------------------------------
nhudson - 2003-05-13 00:23 GMT
-----------------------------------------------------------------------
what are the specs on your computer?
-----------------------------------------------------------------------
sofar - 2003-05-13 21:31 GMT
-----------------------------------------------------------------------
check out disc space on /tmp, /var/spool/lunar, I suspect a disc space
issue to be the cause... please confirm
-----------------------------------------------------------------------
sofar - 2003-05-13 21:32 GMT
-----------------------------------------------------------------------
try with the cache/archive feature OFF to confirm that the problem is in
the achive() code
-----------------------------------------------------------------------
cpowers - 2003-05-14 00:03 GMT
-----------------------------------------------------------------------
This has happened on both of my computers, which are similar but not
identical
home brew machines.
Currently, both use mid-tower ATX cases w/ 250 watt switching power
supply,
ECS K7AMA motherboard, 512 MB PC133 Memory, 40 GB IBM harddrives, lucent
lt winmodems, CD burners and USB wheel mice. Both have NVIDIA GeForce2
MX
video cards (one is a straight MX the other an MX 400). One has an ATHLON
XP1800 cpu, the other a DURON 1200 cpu. Next week they'll be different.
My installation on both machines is similar - a working 4 GB Lunar
partition
mounted on /, with /var/spool/lunar, /var/cache/lunar, /usr/src, and my
home
directories as symbolic links to similar directories on a mounted 20 GB
partition. I do not have the compiler cache turned on - I tried it
initially
but it was eating up all the disk space. There is currently about 1.75
GB
free space on / and lots of room on the common partition.
Under the Options/Features menu, I have selected Archive, Autofix, Color,
Gargage, Preserve, Sustain, Voueur, and Reap. I will try compiling with
the Archive option OFF.
I have used GCC3 from the beginning. Under optimizations, I have
selected
Faster, Athlon XP, NO speed optimizations, NO extra features, x387 fp
math,
Strip, NO addons, 1 cpu, default stack.
I hope this helps. Thanks.
-----------------------------------------------------------------------
nhudson - 2003-05-14 03:55 GMT
-----------------------------------------------------------------------
Your /etc/lunar/local/optimizations file should look similar to this:
PLATFORM=x86
BUILD=i686-pc-linux-gnu
MAKES=
COPT=-pipe
CCOPT=-pipe
BOPT=Fastest
CPU=AthlonXP
SPD=()
STACK=
XTRA=(MMX SSE dnow)
FPM=Both
CC_WARN=(Deprecated)
CCC_WARN=(Deprecated)
LDF=(Strip Combreloc)
ADDON=(CCache)
You said you said 1 for numer of CPU's. This is not used unless you have
more than one so it should be left blank. I had trouble in the past with
compiling because I put 1 there by accedent. Also if your / partition is
only 4GB big and you have /var/spool/lunar in the same partition then it
probobly been pretty much eaten up. My /var/spool/lunar is around 3.5GB
maybe less. Also with all the stuff you have installed to / that can add
up and be causing these problems also. Its not Mozilla I can tell you
that 100% for sure being that noone else has had problems at all. I would
check what I just told you and if that doesnt work I would say either
hardware or memory related probelms are occuring but I am no expert at
that. I will leave the bug open for now for the next few days, but its
not really relavent to mozilla. Oh and another thought Mozilla takes
atleast a good 1 gig space as it compiles and if your 4GB / partition is
pretty full then this would happen also.
-----------------------------------------------------------------------
nhudson - 2003-05-14 03:58 GMT
-----------------------------------------------------------------------
Just to make myself clear on this your MAKES option in your file shouldnt
have athything by it:
MAKES=
instead of :
MAKES=1
HTH
-----------------------------------------------------------------------
cpowers - 2003-05-14 05:06 GMT
-----------------------------------------------------------------------
Turning off the Archive option didn't help. Results were exactly
the same.
I'll try again using your optimizations, but not tonight.
Here's what my optimization file currently looks like on this
computer (I did include the MAKES=1 , will fix that):
PLATFORM=x86
BUILD=i686-pc-linux-gnu
MAKES=1
COPT=-pipe
CCOPT=-pipe
BOPT=Faster
CPU=AthlonXP
SPD=()
STACK=
XTRA=()
FPM=x387
CC_WARN=()
CCC_WARN=()
LDF=(Strip)
ADDON=()
The other computer should be nearly identical.
Disk space is not a problem. The spool, cache, and src directories are
on
a separate partition, now 38 GB in size (I upgraded). / started out at
52% used and finished up the same way - just a few MB added. Likewise,
the
common partition started out at 38% used and finished the same way. I
have
a 1 GB swap partition and top showed less than 1700 KB of it ever used.
Also, please understand, mozilla compiled and installed just fine - I'm
using it right now as shown by this clip from ps:
17151 ? S 0:00 /bin/sh /usr/lib/mozilla-1.3.1/run-mozilla.sh
/usr/lib/mozilla-1.3.1/mozilla-bin
17158 ? S 0:16 /usr/lib/mozilla-1.3.1/mozilla-bin
17160 ? S 0:00 /usr/lib/mozilla-1.3.1/mozilla-bin
17161 ? S 0:00 /usr/lib/mozilla-1.3.1/mozilla-bin
17162 ? S 0:00 /usr/lib/mozilla-1.3.1/mozilla-bin
17163 ? S 0:00 /usr/lib/mozilla-1.3.1/mozilla-bin
17164 ? S 0:00 /usr/lib/mozilla-1.3.1/mozilla-bin
17166 ? S 0:00 /usr/bin/esd -terminate -nobeeps -as 2 -spawnfd
38
It's just the lunar process which hangs and has to be killed, resulting
in the failed install message. Lunar previously completed mozilla-1.3
sucessfully 21 March on this computer and I think 12 April on the other
one, and I don't believe I've touched any of the options until now.
So, again, thanks for looking into this. I'm sure I don't know what's
going on here.
Incidentally, when the install hangs, ps ax shows 3 open
/bin/sh /sbin/lin mozilla processes followed by an md5sum process.
It appears to me that it is hanging in that md5sum process, which is
apparently spawned by the last of the /sbin/lin processes. If I kill
the md5sum process, it comes right back, so I kill the last /sbin/lin
process.
-----------------------------------------------------------------------
cpowers - 2003-05-16 03:31 GMT
-----------------------------------------------------------------------
I give up. I tried your optimizations - no change. I tried reverting to
older lunar's that used to work - no change. So I decided to use the
mozilla
as is and continue. I linned docbook and scrollkeeper as per sofar -
that
worked. Next I started a lunar update and went home. When I got back
today
my 'no rule to make install' problem had bit me again. 20 or so modules
had
sucessfully updated, including gcc and glibc. About an hour later
according
to the log, glibc was compiled again and failed install due to "no rule
to
make install'. About 24 more modules then failed because without glibc,
gcc
couldn't create executeables. So I restored glibc from the last good
cache
and tried to re-lin. Now glibc compiled but failed with the message
cp: /lib/ld-linux..and /lib/ld-lunar... are the same file. I tried to
fix
that and failed.
So, I've downloaded the 1.3.1 iso and started a fresh install on a clean
partition using reiserfs this time. I'll let you know if either problem
bites me again. Thanks.
-----------------------------------------------------------------------
nhudson - 2003-05-16 20:55 GMT
-----------------------------------------------------------------------
Ok I am closing this then ... If you have any more troubles with the new
install please let us know.
More information about the Lunar-bugs
mailing list