Re: assert.m and CTRL-C

[Daniel J Sebald]

On 11/28/2013 05:58 PM, Michael Goffioul wrote:
> Hi,
>
> While testing CTRL-C behavior while running the test suite, I noticed
> strange behavior of assert after interruption. I believe that one of
> them is due to the unprotected usage of call_depth persistent variable
> within assert.m.
>
> If you happen to interrupt octave while executing assert.m, it's
> possible that call_depth is never properly decremented, as expected at
> the end of assert.m. This means that on the next assert call, call_depth
> will never reach -1 and the error will not be triggered. I noticed this
> kind of behavior after interrupting the test suite, something like
> "assert (0,1)" would not trigger any error. Using "clear all" would
> restore assert behavior, which makes me think it's related to the
> persistent variables.
>
> This then brings me to a side question: if assert.m was using
> unwind_protect, would it be robust against CTRL-C interruption? I
> believe it would be, but I'm not 100% sure.
>
> Michael.

I think you are correct Michael.  Could you please copy this post into a 
bug report?

This doesn't seem 3.8 critical.  Breaking out of tests at the wrong time 
seems very likely.  However, there probably aren't a lot of users who 
will be running the test suite.  But this should be a fairly high 
priority item for the development branch once that organization is 
determined.

Thanks,

Dan