Re: Fix nptl/tst-cancel7 for non-bash shells

[David Miller <davem at davemloft dot net>]

From: "Joseph S. Myers" <joseph@codesourcery.com>
Date: Thu, 25 Oct 2012 21:03:21 +0000

> On Thu, 25 Oct 2012, David Miller wrote:
> 
>> For reference, this is the dash fix:
>> 
>> http://git.kernel.org/?p=utils/dash/dash.git;a=commitdiff;h=ee5cbe9fd6bc02f31b4d955606288de36c3d4eab
> 
> That this isn't needed with some dash versions may be relevant 
> information, but it doesn't mean the patch is wrong.  The patch is only 
> wrong, and this is only a dash bug, if POSIX *requires* the implicit exec 
> in this case, rather than it being an optimization, and I didn't see such 
> a requirement in a quick check - if there is such a requirement, what / 
> where is the wording?

I think it's more instructive to list counter examples.

What POSIX shell doesn't handle this properly?

Otherwise it is in fact impossible to cancel out of a system("") call.

And wouldn't we want to test cancellation of system("") using dash if
it is in fact the /bin/sh on the system in question?