Just for completeness sake: David helped me figure out this problem.
It was traced down to a bug in AmberTools 15 that is already fixed in
AmberTools 16.
Applying the following patch solves the issue (vpivot was being used
uninitialized):
--- amber14/AmberTools/src/mdgx/Manual.c.orig 2016-06-10
11:39:47.719576723 +0200
+++ amber14/AmberTools/src/mdgx/Manual.c 2016-06-10
11:41:16.670677715 +0200
.. -142,9 +142,10 ..
endfound = 0;
j = 0;
while (endfound == 0) {
+ vpivot = j + width;
for (i = 0; i < width; i++) {
if (vpar[j+i] == ' ') {
- i++;
+ i++;
vpivot = j+i;
}
if (vpar[j+i] == '\0') {
regards,
Kenneth
On 08/06/16 23:53, David Cerutti wrote:
> This is my program, and while we haven't seen this problem before, it's
> obviously up to me to figure this one out. Let's take this offline and
> I'll see if I can solve the problem.
>
> Dave
>
>
> On Wed, Jun 8, 2016 at 4:28 PM, Kenneth Hoste <kenneth.hoste.ugent.be>
> wrote:
>
>> Hello,
>>
>> I'm trying to figure out a weird problem when installing Amber 14 +
>> AmberTools 15, I hope it looks familiar to someone here who could
>> provide pointers on the possible cause.
>>
>> Basically, I'm seeing "make test" hang as follows:
>>
>> make test
>> \_ /bin/sh ./test_at_serial.sh
>> \_ make -k -f Makefile test.serial
>> | \_ /bin/sh -c cd mdgx/1p7e && ./Test.1p7e
>> | \_ /bin/bash ./Test.1p7e
>> | \_ /<prefix>/software/Amber/14-intel-2016a/bin/mdgx -O
>> -i mdgx.in
>>
>> The mdgx is logging is taking 100% CPU (single-core), and is logging
>> profusely to <prefix>/test/mdgx/1p7e/gb30.out, albeit mostly with
>> newlines...
>>
>> The head of the log file is pasted below, all other lines are empty
>> lines (*lots* of them); this is the size of the log file after 45m:
>>
>> $ wc -l
>> <prefix>/software/Amber/14-intel-2016a/AmberTools/test/mdgx/1p7e/gb30.out
>> 16869489392
>>
>> The '/user/scra' bit at the end of the log file is copy-pasted as it is
>> included in the log file, and is the start of the installation path of
>> Amber.
>> It seems to be cut off for some reason, after which the test goes into
>> lala-land and starts printing newlines.
>>
>> The problem seems to be the length of the path: if I install Amber to a
>> shorter path, the test does not hang.
>>
>> However, the path is not extremely long according to me; the full path
>> to the test/mdgx/1p7e/gb30.out log file is just 151 characters long (all
>> ASCII characters).
>>
>> Does this look familiar to someone? Is there a way to fix it (other than
>> using a shorter install path, obviously)?
>>
>>
>> Contents of test/mdgx/1p7e/gb30.out:
>>
>>
>> <++>-----------------------------------------------------------------------<++>
>> <++> mdgx: A molecular dynamics engine in the AMBER suite of
>> programs <++>
>> <++> <++>
>> <++> Written by David S. Cerutti, Case Group
>> (2009) <++>
>>
>> <++>-----------------------------------------------------------------------<++>
>>
>> Run on Wed Jun 8 15:53:52 2016
>>
>> INPUT FILE ASSIGNMENTS:
>> Designation Base name Suffix
>> ----------------- ---------------------------------------------
>> ----------
>> Input File mdgx.in NONE
>> Input Coordinates inpcrd NONE
>> Topology prmtop NONE
>>
>> OUTPUT FILE ASSIGNMENTS:
>> Designation Base name Suffix
>> ----------------- ---------------------------------------------
>> ----------
>> State Information gb3 .out
>> Restart file gb3 .rst
>>
>> !!! Output overwriting is ENABLED !!!
>>
>> <++>-----------------------------------------------------------------------<++>
>>
>> INPUT LINE TEXT:
>>
>> /user/scra
>>
>>
>>
>>
>>
>> regards,
>>
>> Kenneth
>>
>>
>>
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jun 10 2016 - 04:30:02 PDT
