On Fri, 16 Oct 2015, dev wrote:

> The older SPARC server did the build in twice the time ( yep, clock
> speed isn't everything in the SPARC world ) and the testsuite has been
> running now for 26 hours 52 minutes and 37 seconds fully wedged on test
> 0575 where the last line I see looks like :
>
> test 0574...OK (404 out of 984, remaining: 12:46)
> test 0575...

Does it work if you run only that single test case?

Like this:

$ cd tests
$ ./runtests.pl 575

I can't see what makes 575 particularly sensitive for this platform though.

> So I hit CTRL-C over and over. The things ends like so :
>
> test 0576...FAILED

That just shows we're lousy at handling ctrl-c to end the tests properly. I've
sometimes seen it work if ctrl-c is hit a second time.

> Also all the defunct processes vanish neatly in a puff of smoke.

... at least we clean those up nicely then!

> So I guess the question is, on a slower older system do I need to grant
> special considerations or perhaps indicate the the timeouts on a test should
> be extended or any thoughts ?

Maybe. We need some more details from the failed tests to be able to tell for
sure.

Also, if you want to try it out, you can add '575' to tests/data/DISABLED
before running all the tests as then it'll skip that single test. In case you
want to make sure everything else still runs fine.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html