Re: .DELETE_ON_ERROR & error propagation through multi-command

[Robert]

Paul Smith wrote:
> On Tue, 2011-11-08 at 15:27 +0100, Robert wrote:
> > > I'm not strong in Windows scripting so I can't say the best way to solve
> > > your problem.  Presumably you want to preserve the exit code of the
> > > python script, then do the "cd ..", then exit from the recipe line using
> > > the saved exit code.
> >
> > thanks.  The method with&&  seems to work here on Windows too.
> > MingW GNU make seems to use a sort of MingW sh .
> > Well, I see Windows cmd.exe shell also understands the&&
>
> Just remember that in this case, the final "cd .." is not performed.
>
> That may or may not matter to you.  As I mentioned, in UNIX systems it's
> basically a no-op since the working directory is a property of each
> individual process.  I don't think that's true in Windows, so omitting
> the final "cd .." might cause issues.

I tried:


xtest:
        cd install && grrrrr && cd ..
        echo abcdef >x.txt


C:\devel\prj> make xtest
cd install && grrrrr && cd ..
/bin/sh: grrrrr: command not found
make: *** [xtest] Error 127
C:\devel\prj>



So I think when there is a break with error, its anyway over. And 
the calling shell is as its previous folder.


xtest:
        cd install && echo qwertz>y.txt && cd ..
        echo abcdef >x.txt


..puts the y.txt and x.txt as intended in the different folders


xtest:
        cd install && echo qwertz>y.txt
        echo abcdef >x.txt


.. does the same. So the "logical recipe lines" seem to be 
isolated in sub-shells (not affecting cwd) as well on Windows/MingW



Robert