|
From: Keith R. <kr...@is...> - 2004-08-04 08:44:17
|
Dear Eric, Eric J. Walter wrote: > I am not having this problem, using your ti.param file and doing the > exact steps you say. > > Does anyone else have this problem? > > Could you tell me your machine type, whether you compiled from > source or used an executable and which version of grace you > are using? I compiled from source. I'm using Mandrake linux 10.0 with libc 2.3.3, gcc 3.3.2 and Intel Fortran 8.0. The version of grace is 5.1.14. This is not the problem -- I can see that the grace input file only contains 4 datasets and appears to be complete. > > Furthermore, your log file is not complete. Maybe there is some kind > of buffer flushing problem? Aha. This is a very good clue. I just tested my executable in interactive mode. After the "all" and "plot wp" commands the file ti.plt_nl was of zero length. I then issued the "quit" command whereupon ti.plt_nl received its data. So the root cause is a buffer flush issue. And I see in atom/do_nl.c that the code which writes <>.plt_nl does not have a flcose() call as all the other file writes do. Inserting this at do_nl.c line 255 fixes the problem. I guess I'll also have to find out how to flush the log file stream too. Keith Refson ------------------------------------------------- > > > > > Sometimes I find that Opium fails to plot any pseudowavefunctions after > an apparently successful generation and test. The file ti.plt_nl > is of zero length. There are no apparent failure messages in the log file. > > This happens when I invoke opium interactively > > opium ti.param ti.log > > and issue > > all > plot wp > > interactively but NOT when I run non-interactively, viz > > opium ti.param ti.log all wp > > > > I'm attaching the log and input files in case it's something specific > to my installation.... > > Keith Refson -- Dr Keith Refson, Building R3 Rutherford Appleton Laboratory Chilton Didcot kr AT isise Oxfordshire OX11 0QX DOT nd DOT rl DOT ac DOT uk |