[Swan-dev] sanitize.sh vs re-sanitize.sh (vs swantest.sh)

[Paul Wouters]

On Fri, 20 Mar 2015, Andrew Cagney wrote:

> (I was hoping re-sanitize.sh was a script to update the reference output, little did I know)

We don't have a tool that patches the reference output with the diffs in OUTPUT/

> sanitize.sh
>  - run from swantest
>  - leaves empty diffs when a test passes :-(
>  - doesn't check for crashes and assert failures :-(
> 
> re-sanitize.sh
>  - run from pluto-testlist-scan.sh (nice script)
>  - doesn't leave empty diffs :-)
>  - doesn't check for crashes and assert failures :-(
> 
> swantest
>  - checks for for crashes and assert failures
> 
> Does anyone see a problem with me cherry-picking the best bits:

Note that when changing a sanitizer rule, running sanitize.sh or
re-sanitize.sh allows you to update the output without rerunning the
whole test case. That is useful :P

> - does the sanitize
> - doesn't leave empty diffs
> - does check for core files and assertion failures (those messages can be put into the diff)
> - and exits with 1 if the test doesn't appear to have passed
> 
> perhaps called postmortem.sh

As long as it can run from swantest as well as standalone, I'm fine with
it.

Paul