Re: [Virtuoso-users] VOS 5.0.14 make check fails on 10, passes on 9 Ubuntu

2011-08-04 Thread Dondo, Brian 
# in which I realize I'd forgotten to get the mailing list in the To box
# and only sent the last two to Tim (again)
# a compendium...

Hello again,

I've pretty well given up on 5.0.14 for now.  I'm getting ready to install 
6.1.3 (compiled for x86_64) to production.  In the process I trimmed down the 
options and eased up on the fancy dancing and got it to the point where I can 
run a successful make check WITH the bpel VAD included, which really, really 
puts my mind at ease.  I wasn't happy with just excluding it because it was 
pooching over what looked like a hung instance from a previous test. Couldn't 
really tell.  Now it's about as good as I'm going to get it.

It's worth noting, for those who will listen, that I couldn't do that without 
setting the system locale using update-locale and logging out and back in.  
Just tweaking the shell's environment variables didn't cut it for that.

I've attached a fresh copy of the installation script I used, which includes 
some notes about some of the choices I had to make to get things to check 
clean.  There's probably a few more libraries and dev packs and stuff in there 
than necessary but I know for sure XML and Imagemagick wouldn't grab without 
their's.

The thing about locale leads to this question:

Is there any way to confirm whether or not running the binaries in a en_CA.utf8 
environment will be an issue or is it advisable to keep the locale set to 
C/POSIX for runtime?

Another question: The Debian packager "checkinstall" is noticing that /binsrc/samples/demo/install_ini.sh is supposed to install more or 
less in place, branching off  rather than within the Debian 
layout.  Do I really need that script where it wants to go or can I slide it 
out of the way so it's not flapping around on a thin branch of the tree and put 
it somewhere accessible later if I need it?

Thanks,
b

# and

Sorry for the three-in-a-row, but here's the kicker.

In an effort to answer my own question about utf8, because that is after all 
what we do for a living, I took the following steps.

1) Kept the login shell locale at C/POSIX and ran the make
2) sudo update-locale LANG=en_CA.utf8 LC_ALL=en_CA.utf8
3) log out and in so the login shell locale is en_CA.utf8
3) trimmed down my environment variable exports to CFLAGS and CXXFLAGS, leaving 
the install shell locale at en_CA.utf8.  That way the test script shells would 
set to C/POSIX.
4) ran a make check

It worked.  Clean as a whistle. I couldn't compile and make check with 
--enable-bpel-vad without the login shell starting in C/POSIX, but I can make 
check with the login shell @ utf8 as long as the make login shell started in 
C/POSIX.  I'm no expert when it comes to compiling but to me that points to a 
problem during compile/link the make process that isn't being trapped.

Thought I'd mention it.

# still wondering about install_ini.sh
# of course, I haven't even looked at it or anything responsible like that.





install.sh
Description: install.sh

Re: [Virtuoso-users] VOS 5.0.14 make check fails on 10, passes on 9 Ubuntu

2011-07-26 Thread Dondo, Brian 
Hello again,

Not sure how the use the mailing list, really, so I hit "reply to all" over 
here in Microsoft world hoping this gets where its supposed to go without 
turning into spam.

I've attached two versions of the results of:

egrep -l '^(\*\*\*.*FAILED|\*\*\*.*ABORTED)' $(find ./ -name "*" | egrep 
".+output$|.+log$")

from the build directory so you're only getting the *.outputs with \n***FAILED 
in them.  At least I THINK that's what the ^ does.  I stole the pattern from 
you guys so don't give me grief it misses something. ;-))

Now if Symantec doesn't think they've got viruses I'm pretty sure they'll make 
it past server support's Megabyte Police.  My apologies if they don't.

The 'crash' version is the one that grinds to a halt.  The 'graceful' at least 
makes it back to the bash prompt and tells me to contact support.  You'll 
notice in the (path removed) install script there's a couple of tweaks that I 
stole or adapted from the source version of the Ubuntu package.  It doesn't 
finish but at least it makes it back to a prompt.

Neither version of the script works on Ubuntu 10.04 with a 2.6.32.28 kernel.

Thanks for you help +| reason for hope.

b


-Original Message-
From: Tim Haynes [mailto:tim.hay...@openlinksw.co.uk] 
Sent: Friday, July 22, 2011 11:16 AM
To: virtuoso-users@lists.sourceforge.net
Subject: Re: [Virtuoso-users] VOS 5.0.14 make check fails on 10, passes on 9 
Ubuntu

On 21/07/2011 21:36, Dondo, Brian wrote:
> Hi.  (am I supposed to say hi on mailing lists?)

You can say hi if you want ;)

> Building 6.1.3 and 5.0.14 VSO on Ubuntu 9.10 and 10.04.  Builds on 32 bit 
> Ubuntu presents no problems with either release of VSO.
> 
> It's worth mentioning right away I'm using the -with-iodbc=DIR option.  (=DIR 
> because for some reason in 10.04 the make isn't able to figure the absolute 
> path when it needs it)

That's ok, as long as you have libiodbc2-dev installed on all platforms
involved.

> However, we're planning on going live using 64-bit compilations on 64-bit 
> servers.  The other combinations were done because I'm stuck working the 
> angles on this.
> 
> The targeted combination is 5.0.14 VSO on 10.04 Ubuntu 64-bit.  This is where 
> it gets weird.  I can make and make check 5.0.14 on Ubuntu 9.10 64-bit with 
> no problems BUT on 10.04 it compiles clean but I can't run a successful make 
> check.  There're a couple flashes of errors near the beginning of the check 
> and then things go well until it eventually grinds to a halt(ish) when 
> "Running a subset of TPC-D queries against "
> 
> And this is where it gets weirder still.  If I run the make on 10.04 64-bit, 
> tarball the make directory over to a 9.10 64-bit OS and run the make check 
> there, it runs clean!
> 
> Is there a quirk in Ubuntu that could be causing this?  Is it the test suite 
> on 10.04?  The 10.04 compile tests clean on 9.10. Can I trust the binaries on 
> 10.04 if I package them up and install them using Debian utilities?

It's certainly possible there might be quirks; could be that you have older
versions of required packages[0]; could be locale settings[1] at fault
(gawk has been known to be problematic on utf-8 locates before now).

[0] Please double-check against
<http://virtuoso.openlinksw.com/dataspace/dav/wiki/Main/VOSMake#Package%20Dependencies>

[1] Please check the output of `locale` on all environments

Best bet is if you send us the various tsuite log files, mostly
binsrc/tests/suite/testall*{output,log} so we can look for more precise errors.

HTH,

~Tim
-- 
Tim Haynes
Product Development Consultant
OpenLink Software
<http://www.openlinksw.com/>
<http://twitter.com/openlink>

--
10 Tips for Better Web Security
Learn 10 ways to better secure your business today. Topics covered include:
Web security, SSL, hacker attacks & Denial of Service (DoS), private keys,
security Microsoft Exchange, secure Instant Messaging, and much more.
http://www.accelacomm.com/jaw/sfnl/114/51426210/
___
Virtuoso-users mailing list
Virtuoso-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/virtuoso-users


lucid64-2.6.32-32-server-VSO-5.0.14-graceful_fail.tar.gz
Description: lucid64-2.6.32-32-server-VSO-5.0.14-graceful_fail.tar.gz


lucid64-2.6.32-32-server-VSO-5.0.14-crash_check.tar.gz
Description: lucid64-2.6.32-32-server-VSO-5.0.14-crash_check.tar.gz

Re: [Virtuoso-users] VOS 5.0.14 make check fails on 10, passes on 9 Ubuntu

2011-07-22 Thread Tim Haynes 
On 21/07/2011 21:36, Dondo, Brian wrote:
> Hi.  (am I supposed to say hi on mailing lists?)

You can say hi if you want ;)

> Building 6.1.3 and 5.0.14 VSO on Ubuntu 9.10 and 10.04.  Builds on 32 bit 
> Ubuntu presents no problems with either release of VSO.
> 
> It's worth mentioning right away I'm using the -with-iodbc=DIR option.  (=DIR 
> because for some reason in 10.04 the make isn't able to figure the absolute 
> path when it needs it)

That's ok, as long as you have libiodbc2-dev installed on all platforms
involved.

> However, we're planning on going live using 64-bit compilations on 64-bit 
> servers.  The other combinations were done because I'm stuck working the 
> angles on this.
> 
> The targeted combination is 5.0.14 VSO on 10.04 Ubuntu 64-bit.  This is where 
> it gets weird.  I can make and make check 5.0.14 on Ubuntu 9.10 64-bit with 
> no problems BUT on 10.04 it compiles clean but I can't run a successful make 
> check.  There're a couple flashes of errors near the beginning of the check 
> and then things go well until it eventually grinds to a halt(ish) when 
> "Running a subset of TPC-D queries against "
> 
> And this is where it gets weirder still.  If I run the make on 10.04 64-bit, 
> tarball the make directory over to a 9.10 64-bit OS and run the make check 
> there, it runs clean!
> 
> Is there a quirk in Ubuntu that could be causing this?  Is it the test suite 
> on 10.04?  The 10.04 compile tests clean on 9.10. Can I trust the binaries on 
> 10.04 if I package them up and install them using Debian utilities?

It's certainly possible there might be quirks; could be that you have older
versions of required packages[0]; could be locale settings[1] at fault
(gawk has been known to be problematic on utf-8 locates before now).

[0] Please double-check against


[1] Please check the output of `locale` on all environments

Best bet is if you send us the various tsuite log files, mostly
binsrc/tests/suite/testall*{output,log} so we can look for more precise errors.

HTH,

~Tim
-- 
Tim Haynes
Product Development Consultant
OpenLink Software