Hi Ben,
thank you for the work you've done. I am not the bondtype author,
but enjoyed playing with it and hence the curiosity regarding the
possible bug. Could you please 'cvs update -d' in src/antechamber,
then 'make bondtype.orig' and try it instead of the 'bondtype'
[just to rule out few more possibilities].
Best,
Volodymyr
On Wed, November 25, 2009 16:59, Ben Roberts wrote:
> All:
>
> It seems that something is going very wrong inside bondtype.c on
> chemokine.
>
> In my case, the fatal error is triggered because there exists a bond for
> which jflag is set to 1 on Snow Leopard, compared to -1 for Linux. This
> causes the code to travel down a different path, where it checks for the
> frozen atom type, discovers that it's 0, and exits (relatively)
> gracefully.
>
> However, that's not the only discrepancy between Linux and Snow Leopard.
> Many of the entries for parts of BOND data structures (type, type2,
> jflag, etc.) are different. Sometimes, on Linux, the type entry for a
> certain bond will contain a large negative number (in the millions or
> higher), while on Snow Leopard the type entry for the same bond will be
> 0. Other times the difference is more subtle (e.g., -2 instead of 0, or
> the flag case above).
>
> I have put a bunch of files up on a web server, rather than trying to
> send them to the list, since some of them are large. I hope this is OK.
> If it's a problem, let me know and I can take them down.
>
> * bondtype.c - modified by turning debug on at the top of the file, and
> added a bunch of print statements.
>
> * 1b2t.csh - a modified version of Run.1b2t, but things are appended to
> "antechamber.out" rather than overwriting it.
>
> * and two text files (originally called antechamber.out, and produced by
> 1b2t.csh) - one generated on a Linux box, and the other on Snow Leopard.
> Which is which should be obvious from the names of the files.
>
> I hope that helps people figure out what's going on here, and that it's
> not another NAB-like problem. Those who are working on this, let me know
> if you would like more information or a trial run of something.
>
> Ben
>
> On Tue, 2009-11-24 at 11:36 -0500, Volodymyr Babin wrote:
>> Hi everyone,
>>
>> I don't have the OS around unfortunately to try it myself.
>> You could try to run [change -pf yes to -pf no first]
>>
>> /Applications/Amber11/intel-mkl/bin/bondtype -j full \
>> -i ANTECHAMBER_BOND_TYPE.AC0 -o ANTECHAMBER_BOND_TYPE.AC -f ac
>>
>> manually and examine the output if it is wrong indeed (to rule out
>> some parsing errors). You could then try to nail it down further
>> by injecting 'printf()' statements around. Alternatively, I could
>> take a look if someone could give me a (temporary) shell at a leopard.
>>
>> Best,
>>
>> Volodymyr
>>
>> On Tue, November 24, 2009 11:27, Ben Roberts wrote:
>> > Hi Scott,
>> >
>> > Last I knew, this problem still exists. It seems bondtype() was
>> > returning the wrong values for certain things.
>> >
>> > Specifically:
>> >
>> > <snip>
>> >
>> > ---Judge bond type for Residue 12 with ID of 11 and Name of THR ---
>> >
>> > The frozen atom type can only be 1, 2, 3, 7 (aromatic single), 8
>> > (aromatic double)
>> > Error: cannot run "/Applications/Amber11/intel-mkl/bin/bondtype -j
>> full
>> > -i ANTECHAMBER_BOND_TYPE.AC0 -o ANTECHAMBER_BOND_TYPE.AC -f ac" in
>> > judgebondtype() of antechamber.c properly, exit
>> >
>> > Interestingly enough, it grumbles at a different point to Alan's
>> > original bug report.
>> >
>> > I'll see what I can find. Is there any indication as to what,
>> > technically speaking, is causing the problem?
>> >
>> > Ben
>> >
>> > Scott Brozell wrote:
>> >> Hi,
>> >>
>> >> What is the status of the antechamber tests on Uncia uncia ?
>> >> In particular, does the chemokine problem at the tail below still
>> exist
>> >> ?
>> >>
>> >> thanks,
>> >> Scott
>
>
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Nov 25 2009 - 19:00:03 PST
