[Dock-fans] Check_serial.out

[Scott Brozell]

Hi,

On Thu, 8 May 2008, Francesco Pietra wrote:

> I should have better explained what "lenny" (Debian Linux) means in present context.
> > I recently upgraded Debian Linux amd64 from "etch" (the stable distribution) to "lenny" (the intermediate distribution) as I used in the past the intermediate distribution (at that time it was "etch"). Then, I applied the bug fixes (I should say carefully so that there should be no misprint) and re-compiled on the former install directory, i.e., overwriting the executables (after running "./configure gnu" for serial and "./configure
gnu parallel" for parallel). I saved a copy of former compilation, renamed dock6_old.
>

On Fri, 9 May 2008, Francesco Pietra wrote:
> --- On Thu, 5/8/08, Scott Brozell <sbrozell at scripps.edu> wrote:
> 
> > Ok, your  system upgrade probably explains all the dock test diffs.
> 
> I have something new, bad or good as it might be.
> 
> I downloaded DOCK6.2 again, applied the already elaborated bugfix 1 and 2, and compiled serial on the same Supermicro/debian lenny system below (to rule out any trivial mistake from my previous doing).
> 
> Then I examined the dock6.2 compilation of 13 Mar 2008, currently renamed old_dock6. At that time the system was a couple of dual-opterons on a Tyan motherboard under Debian amd64 etch (stable) and there were no bugfixes. The two raid1 HDDs from this system were simply moved to the new system Supermicro motherboard with four dual-opteron cpus. RAM from Tyan (Kingston ECC 400MHz) was also moved to Supermicro.
> 
> Well, both the good agreement and the large diff you remarked below as unrelated to bugfixes:
...
> are exactly the same in both instances, the new compilation from scratch and the old one on a different motherboard. I apologize for having overlooked such diff in the 13 Mar 2008 compilation. Though, it is not the new hardware/OS that has introduced that problem. Assuming that other users were not faced by such issues, at the moment I can't imagine what best to do. As to the Debian list, so far I only posed a question about the two installations of gnu; all other issues could not be understood.
> 

Ok, platform lenny and platform old_dock6 yield the same test results.
This is good because bugfixes 1 and 2 do not and should not affect the results.
This also suggests that the test suite is more robust than advertised.

An important problem with the test suite may just be dated .save files.
The .save files are from an ia32 
Linux piv-login1 2.6.18.1 #12 SMP Sun Feb 10 17:08:11 EST 2008 i686 i686 i386 GNU/Linux
gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-53)
mostly because we still have not fixed all 64 bit issues.
So your diffs may be due to differences across these platforms,
primarily from binary files, floating point arithmetic, etc.

Scott