Re: testsuite failures when test scripts are run with zsh

[Stefano Lattarini]

At Tuesday 08 December 2009, Ralf Wildenhues <address@hidden> wrote:
> Hi Stefano,
> 
> Thanks for continuously updating your patch; and sorry for not
>  getting it in before 1.11.1/1.10.3.
Don't worry, I see that there are more important patches pending and
more urgent work to be done.

Anyway, in the meantime I had taken some further steps, and created
a series of three patches (my original one, plus other releated two),
which I will soon post in a new thread.  Hope you are OK with this.

>  Your patch requires some portability fixes and some testing on
>  several systems, and for the releases I decided to play safe and
>  worry more about not getting any regressions in testing
>  functionality
That's a good policy, and I fully agree with it.
>  than getting us to work well with another shell (which, almost by
>  definition, isn't the only one on the system; if it were, you could
>  likely invoke zsh as /bin/sh and it would work well with the
>  current code).
>
Mmhh... it seems that you are confusing the issue tackled by this
patch with a different, minor problem due to Zsh redefining `$0' in
sourced scripts (but your conclusion is mostly correct anyway, as I'll
tell later).

The problem with `$0' is admittedly a minor one, and a very simple
workaround for it is already described in `tests/README' (changes
introduced in commits 53fe020252642f4b923b8c77392e37e4c5b7038d and
674ee2f48686aa67507f9fa6dd67487f8a0e3b93 on master branch).  There is
also a different patch of mine (still pending) which automatically enables
a workaround for this problem (Zsh 4.x only); it's posted in the thread
at the following address (please don't stop at the first message, as I
made a couple of ugly mistakes in my first two attempts):
 http://lists.gnu.org/archive/html/automake-patches/2009-11/msg00003.html

The issue the current patch tries to address is described in details
in this older post:
 http://lists.gnu.org/archive/html/bug-automake/2009-08/msg00020.html

You can see that it's quite a different issue.  But the good thing is
that it shouldn't occur when /bin/sh is zsh, because (cut&paste from
the linked post):
  "... the only systems I heard of that use zsh as /bin/sh
   were MacOS X 10.0 or 10.1, and google tells me that they used
   a 3.0.something version of zsh, which does not show the
  aforementioned bug."

> I will try to work on it soon though.
Take all the time you need.

Thanks,
    Stefano