Detecting errors in a pipe in ash

[Denys Vlasenko]

On Friday 04 December 2009 12:37, Michael Abbott wrote:
> On Fri, 4 Dec 2009, Michael Abbott wrote:
> 
> > When running the command:
> > 	command1 | command2
> > is there any way to detect in ash that command1 has failed?  
> > Unfortunately the default return code is set to that of the last command 
> > in the pipe, command2 in this case.
> > 
> > In bash I can run
> > 	set -o pipefail
> > which sets the error code to that of the last non-zero status in the 
> > pipe, which is normally precisely what it wanted.  Alternatively, of 
> > course, I can always interrogate $PIPESTATUS.
> > 
> > Unfortunately ash understands neither of these constructs.  Is there a 
> > third workaround, or do I need to figure out how to hack ash?  
> > Fortunately it looks as if `set -o` is already implemented, so I may be 
> > in luck...
> 
> Well, looking more closely at ash it's clear that the only solution is to 
> implement this function.  Fortunately (unless I've missed something, 
> likely enough, there's a lot of code there) it seems easier than I feared:

Applied, thanks!

I edited it a bit and made conditional on ASH_BASH_COMPAT.

Try current git.
--
vda