Re: [lmi] GUI test: 'validate_output_census' failure

[Greg Chicares]

On 2015-12-14 02:49, Vadim Zeitlin wrote:
> On Sun, 13 Dec 2015 20:57:09 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> Did we change the program's behavior without updating the tests?
> 
>  I'm afraid we did, in r6323:
> 
> http://svn.savannah.nongnu.org/viewvc?view=rev&root=lmi&revision=6323

Committed 20151214T1847Z, revision 6444.

>  I attach the patch with the changes that make the pass for me again, but I
> didn't have time to test them with MinGW yet, only MSVC. They shouldn't be
> compiler-specific however, so I'm relatively confident that they should
> work for you too.

Tested and confirmed with MinGW gcc.

And here's the five-why's analysis:

>  The real problem is that the test is not being run automatically. It would
> be really nice to use a CI with lmi. As always, nothing is fool proof but
> the likelihood of a test remaining broken for so long would be much smaller
> if we had automatic checks for it. I'd really like to set something up for
> lmi.

Look for "nychthemeral" in 'GNUmakefile' to see my ideas.

Soon I'll have a spare supermicro server that we can set up to do anything
you like. We can rebuild wx from trunk every hour, for example, and lmi
along with it. What we most want to test is our msw distribution; I don't
know how nicely 'cron' penetrates VMs, but perhaps that's all the more
reason to try running (GUI tests, especially) under 'wine'.