Re: [PATCH v3] [BZ #14579] rtld: limit self loading check to normalmode only

[Carlos O'Donell <carlos at systemhalted dot org>]

On Sun, Sep 16, 2012 at 03:03:34AM +0400, Dmitry V. Levin wrote:
> I'm not quite happy to deviate from the current practice with my first
> commit.  It would make a bad precedent. :)

No, it makes a *good* precedent. New changes start off on the right
foot, old code remains as is, incremental progress.
 
> If there is a strong preference to use 'set -e' inside test scripts,
> lets update all these cases altogether.  

No. That's a distinct patch from this patch. I would be excited to see
you post a second patch fixing all of the `-e' behaviour :-)

> > > set -e makes the script exit with a non-zero status if something goes
> > > wrong.  Not the program being tested but something else.  I think
> > > it is a good default behavior for test scripts.
> > 
> > I know what it does.  I can't see how it would make any possible difference
> > in your script, which doesn't run anything other than the tested program.
> 
> The script is quite simple now, but who knows whether it will always
> remain that simple.  One may later add a test for rtld output, which would
> likely run an external executable like grep or cat.  I prefer to be on the
> safe side and make it 'set -e'-agnostic from beginning.

I agree that `set -e' should be in the script, and personally I think we
should be using that for *all* of our scripts regardless of how complex
they may or may not be, it's just good policy. I know that using `set
-e' may sometimes complicate the invocations in a script, but that just
means we need to develop some more robust common script functions that
other scripts can use.

Does that make sense?

Cheers,
Carlos.