Hi,
On branch tleap-f-abortonerror, I have added the basic functionality.
I recommend that we merge this into the master after the release and
give it a thorough test drive before releasing it. (We do not know
how much we are relying on the old behavior).
Presently, there is no control switch - it is always on. It might be
useful to have a command line flag such as -a and/or a command such
as abortOnError. And currently, the only error that can cause an
abort is failing to open a file.
Here are some details:
fBasicsMyFopen aborts on error when commands are sourced from a file.
Presently this applies to commands inside a sourced file, ie inside bla for
-f bla, but it does not apply to the file itself, ie, if bla cannot be opened
then there is no abort; nor does it apply to the default startup leaprc file.
The isatty approach does not work because leap is
always running interactively, ie, stdin is a tty even when files are
being sourced.
scott
On Tue, Feb 21, 2017 at 03:27:40AM -0500, Scott Brozell wrote:
> Ok, I'll investigate.
>
> On Mon, Feb 20, 2017 at 04:38:17PM -0500, Hai Nguyen wrote:
> > +1 for exiting right away if getting error since the final prmtop is NOT ok
> > to use.
> >
> > On Mon, Feb 20, 2017 at 4:32 PM, Justin MacCallum <justin.maccallum.me.com>
> > >
> > > We routinely use tleap non-interactively, and this behaviour bites us
> > > occasionally. It would be great if there was a mechanism to specify that
> > > tleap should die with a suitable error code whenever ti would otherwise
> > > print an error when called interactively.
>
> On Mon, Feb 20, 2017 at 11:48:10AM -0500, David Case wrote:
> > This is a long-standing behavior for tleap:
> >
> > > MSE = loadmol2 MSE.mol2
> > Could not open file MSE.mol2: not found
> >
> > ...soldier on, continue with the next command.
> >
> > The above behavior is appropriate for interactive use, to give the user
> > the chance to correct a typing mistake. But it mainly causes confusion
> > for non-interactive use, since the error message that almost invevitably
> > will come often hides the true source of the error.
> >
> > Could we not use isatty() to make "not found" a fatal error for
> > non-interactive use? Can some kind soul with spare time try to do this?
> >
> > [I know that there are some workflows that rely on the current behavior,
> > e.g. they load files that may not exist but also may not be needed.
> > How much should we worry about breaking such behavior? Do we need a
> > command-line switch (or a new "global" variable) to control whether
> > "non-found" is fatal?]
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Sun Mar 26 2017 - 16:00:02 PDT
