Re: [DejaGnu] handling of errors in tests scripts

[Rob Savoye]

On Tue, May 08, 2001 at 05:18:53PM +0200, Martin Muenstermann wrote:

> We found it difficult to create bullet proof test scripts, i.e. in
> certain (error) conditions the scripts generate a Tcl error (invalid
> spawn_id etc.pp.). Unfortunately in these cases "make check" generates
> no error state, so the batch build is continued in spite of the error
> during "make check".
> 
> IMHO it would be nicer if dejagnu returned an error exit state in these
> cases. This could be done quiet simple by adding the lines
> ################
>     global exit_status
>     set exit_status 1
> ################
> to file runtest.exp in the catch block of 
> catch "uplevel #0 source $test_file_name"
> (line 1433 in V1.4.1).

  diff -u is a little easier. :-) I thought about it, and while I agree
with, you, I'm not 100% sure this is the best behavior. At Cygnus, we
used to prefer our batch jobs not stop if there was a bug in the Tcl
code of the test suite. I might make it a command line option, since
I'm not sure one behaviour is any different from the other, depending
on personal preference. I'd discuss it on this list, but I don't think 
there are enough people subscribed yet to have a real discussion. 

        - rob -