path not working when executing from script...

English Applications
#1

Hello,
Upon upgrading to suse11.2, I’ve noticed a reoccurring problem.

I use tcsh. When I execute by hand a command that is in my path…located in say “~/bin/myscript”…the command executes without an issue. I don’t need to use the absolute path “~/bin/myscript”, I can simply type the command “myscript” since the path “~/bin” is in my .tcshrc file.

When I execute the same command from a script, I get a “command not found” error. A similar issue is when I use the absolute command, but a library necessary the programs’ execution is defined within my .tcshrc file. I’ll get a library not found error.

I didn’t have this issue with earlier versions of suse.

Thanks for any help,
Paul

#2

Additional info that I just realized is relevant:
The system is a cluster with nodes nfs-mounting user directories from the server’s /home/ directory. I submit jobs via a ssh command such as " ssh mymachine myscript".

When I run "ssh mymachine ‘which ls’ " I get back the path “/bin/ls” like I should. If I run "ssh mymachine ‘which myscript’ ", I get a command not found error. If I run “which myscript” I get the correct path.

As mentioned, all machines use the same .tcshrc file because they all share the same nfs mounted directory.

#3

Reading* man ssh* I found following:

If command is specified, it is executed on the remote host instead of a login shell.

which is about the command parameter in your ssh call (the “which ls” or the name of your executable script).
and:

PATH
Set to the default PATH as specified when compiling ssh .

and I doubt that ssh is compiled while* ~/bin/myscript* was one of the items in the PATH environment variable.

And when it executes that command it nowhere says that it does so using any shell at all.

#4

Thanks for the reply.

I have been running the same scripts since suse10.0 and have only recently had this issue.

Short of reinstalling ssh on all my nodes, do you have a suggested fix.

Right now, as a work around, I use a wrapper script with a explicitly defined shell.

#5

I can not tell why it worked in 10.0 where it shoud not work also :slight_smile:

For a solution look in that man page. Browse a bit, there are several places where adding something to a PATH is mentioned, but I do not think I should read them all where you can do this yourself. Also I do not use tcsh here for testing.

#6

As a follow-up, this was my eventual solution.

I created a wrapper script that explicitly defined the shell I needed. So instead of executing “ssh myscript”, I’d make a wrapper script “mywrapper” that has the shell information necessary for my target executable.

mywrapper:
#!/bin/tcsh
myscript

So my final execution is “ssh ~/pathtofile/mywrapper”

Still a mystery why this seemed unnecessary for previous releases.