Re: [jamboost] Re: V2 testing

From: Vincent Jamart (vincent.jamart_at_[hidden])
Date: 2004-09-21 05:55:49

On Tue, 2004-09-21 at 11:45, Vladimir Prus wrote:
> Vincent Jamart wrote:
(...)
> There are some problems:
> > -AIX 5.1/vacpp 6.0 (64bit)
> > -Irix 6.5/MiprsPRO 7.4 (64bit)
> 
> The logs appear to be truncated. Maybe, it's some glitch on the server, or in 
> Python on those platforms?
> 
I guess it's a python issue, because when I run it by hand, on AIX I
got:

python test_all.py
skipping railsys test since QTDIR environment variable is unset
unit_tests :
Traceback (most recent call last):
File "test_all.py", line 140, in ?
run_tests(critical_tests, tests)
File "test_all.py", line 33, in run_tests
__import__(i)
File "/home/qa/cvstop/boost/tools/build/v2/test/unit_tests.py", line
5, in ?
t = Tester(pass_toolset=0)
File "/home/qa/cvstop/boost/tools/build/v2/test/BoostBuild.py", line
102, in __init__
raise "Don't know directory where jam is build for this system: " +
os.name + "/" + os.uname()[0]
Don't know directory where jam is build for this system: posix/AIX

on Irix:

python test_all.py
skipping railsys test since QTDIR environment variable is unset
unit_tests :
Traceback (most recent call last):
File "test_all.py", line 140, in ?
run_tests(critical_tests, tests)
File "test_all.py", line 33, in run_tests
__import__(i)
File "/disk2/home/qa/cvstop/boost/tools/build/v2/test/unit_tests.py",
line 5, in ?
t = Tester(pass_toolset=0)
File "/disk2/home/qa/cvstop/boost/tools/build/v2/test/BoostBuild.py",
line 102, in __init__
raise "Don't know directory where jam is build for this system: " +
os.name + "/" + os.uname()[0]
Don't know directory where jam is build for this system: posix/IRIX64

These errors are not raised on other platforms with Python 2.3.3 or
2.3.4 (AIX/IRIX have 2.3.3)

> > -AthlonMP1900+(x86-32) GNU/Linux 2.4.18-smp (RedHat 8.0)/icc 8.0
> 
> The test result indicate that gcc is being used, not icc.

I guess the test_all script doesn't read my user-config.jam and as
cc=gcc is default on RH8, I now changed the .bashrc of qa user to use
alias cc=icc before running test_all. Hope this will force icc to be
used on this RT at the next run.
.
> 
> > -Athlon(x86-32) GNU/Linux 2.4.21 (SuSE 9.0)/gcc 3.3.1
> 
> Some of the test failures say that you did not configure "icc" correctly on 
> that box. Anyway, it's not a big problem.
> 
Yes, I saw it this morning. I had both gcc & intel-linux in my
user-config.jam but I commented-out intel-linux as I removed the intel
compilo from my host, it seems that the # is innefective so I removed
the whole line "using intel-linux" below "using gcc". Let's see at the
next run.

> > -Solaris9/gcc 3.3.1 (64bit)
> 
> Here, there's a lot of errors like:
> 
> ld: fatal: file bin/gcc/debug/a.o: wrong ELF class: ELFCLASS64
> 
> What could be wrong? Does ld needs some switch to operate in 64-bit mode?
> 
I don't know. On Solaris I use a site-config.jam <> than on other
systems like you explained me in a former mail in order to set gcc
options in 64-bit mode (using gcc : : : -m64 -mcpu=v9 ;). The flags were
reported to work while building a project using boost but I stil have to
analyse the full build log in order to find similar ELF errors within.

> Thanks,
> Volodya
> 

You're welcome ;-)

----------------------------------------------------
Vincent Jamart - Free Field Technologies
Quality & Systems manager
16 place de l'Universite
B-1348 Louvain-la-Neuve - BELGIUM
Tel: [phone] Fax: [phone]
Mobile: [phone]
mailto:vincent.jamart_at_[hidden] http://www.fft.be
----------------------------------------------------