[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: AT_TESTED and (non)fatal errors
From: |
Noah Misch |
Subject: |
Re: AT_TESTED and (non)fatal errors |
Date: |
Tue, 15 Mar 2005 18:40:27 -0800 |
User-agent: |
Mutt/1.5.5.1i |
On Fri, Mar 11, 2005 at 03:30:39PM +0100, Ralf Wildenhues wrote:
> I would like to run the Autotest testsuite of Libtool HEAD on systems
> which do not have all AT_TESTED() programs available, or possibly under
> a different name.
>
> Right now, the testsuite fails when it there is at least one program
> missing which is necessary for one test, even if all I want is to run
> some other test which does not need the program. Ideally, I would like
> something like SKIP if the program is not available, and AT_TESTED to
> allow to work more test-local.
That sounds appropriate. Do you have an implementation in mind? If not, I'll
make a note to work something up.
> Furthermore, I would like the names to be overridable, as in
> AUTORECONF=autoreconf259
I suggest `: ${AUTORECONF=autoreconf}' at the beginning of each test group that
uses autoreconf. Use `$AUTORECONF' wherever you formerly used `autoreconf'. A
user can then `./testsuite AUTORECONF=foo' to override the default.
> What is the best way around this, short of creating stub commands in
> $PATH somewhere (which I don't think I can force users to do)?
> Or should we just ditch AT_TESTED and write our own little macro?
There is no good way to continue using AT_TESTED with a scheme like this;
arguments to AT_TESTED are single-quoted in `testsuite', so variable arguments
do not work. I will fix that, but I suggest avoiding AT_TESTED for now.