Bug #36554 for URI: t/heuristic.t : failure during 'make test'

On Sat Jun 07 17:42:55 2008, JKEENAN wrote: Show quoted text
The test fails because it searches for http://www.perl.com/camel.gif, which is no longer found on that page.

On Mon Jun 09 02:20:29 2008, GAAS wrote: Show quoted text
I reported exactly what I got. If it had successfully resolved, the test would have passed and I would never have opened the bug ticket in the first place. Note that it is failing test #6 (MYCOUNTRY = 'bv') but passing in #7 (MYCOUNTRY = 'uk'. (Also, if this test had been written with Test::More descriptions, it would have been easier to diagnose.) Thanks for looking into this. Jim Keenan

On Wed Oct 01 08:04:36 2008, SREZIC wrote: Show quoted text
Right now I can reproduce the failure: $ perl -Mblib -MURI::Heuristic=uf_urlstr -e '$URI::Heuristic::MY_COUNTRY = "bv"; warn uf_urlstr("perl/camel.gif")' http://www.perl.bv/camel.gif at -e line 1. And really, on my machine www.perl.bv seems to exist: $ host www.perl.bv www.perl.bv has address 64.236.172.120 Host www.perl.bv not found: 3(NXDOMAIN) On another machine it does not resolve: $ host www.perl.bv Host www.perl.bv not found: 3(NXDOMAIN) If I type in the URL in my browser, then I get a catchall page from my internet provider. Argh! They are doing nasty things! Regards, Slaven

Encountered this bug this morning. Isn't this just a case of setting the country code to blank in heuristic.t or alternatively altering the regex line to check for .bv instead of the .com|.org as it cuurently seems to. Easy fix surely.

On Sat Jun 07 17:42:55 2008, JKEENAN wrote: Show quoted text
Test #7 passes for reasons that will probably cause trouble in the future. $URI::Heuristic::MY_COUNTRY = "uk"; print "not " unless uf_urlstr("perl/camel.gif") =~ m,^http://www\.perl\.(org|co)\.uk/camel\.gif$,; print "ok 7\n"; http://www.perl.co.uk/camel.gif takes you to what appears to be a domain squatter. http://www.perl.org.uk/camel.gif doesn't resolve. But http://www.perl.org.uk takes you to a site being squatted by a tech company called SCINETIQ.

On Thu Jul 02 20:35:00 2009, todd_rinaldo wrote: Show quoted text
Got this as well today. Strangely it seemed fine the first time or two I installed it, but have been testing a deployment script and wiping it clean after each go. Suddenly I got this error. Now, this particular error appears to be the cause of most of the cpan failures reported, and considering how many other packages depend on URI now would be great if we could do something... I'm not sure exactly what this test is trying to do, but it's reliance of some dubious hosts names is problematic. I guess when this test was written the author assumed 'www.netscape.com' was pretty safe. I'd say change it to www.google.com, but that would mess up since so many countries block google from time to time. I don't completely understand what this test is hoping to achieve, but it seems totally wrong to say: $URI::Heuristic::MY_COUNTRY = "bv"; and then try to match on a .com or something. Maybe .bv was handled differently all those years ago. The docs say something about if the country is two letters, assume is the default country, but that is not clear at all to me what it means. We should probably just skip this test for now since I don't see what is going to get hurt. My guess is all the passes have more to do with netscape.com being unresolvable for most people. Or we can set the dns lookup to an ENV var and skip the test if it's not present. That way the authors can set it for developing if there is some logic there which is important to fix.

Sorry for spamming the bug queue, but I don't see any other place to discuss this. So basically if you turn on debugging, the problem becomes clear: ----- john-napiorkowskis-computer:~ jjn1056$ perl -e 'use URI::Heuristic qw(uf_urlstr); $URI::Heuristic::MY_COUNTRY = "bv"; $URI::Heuristic::DEBUG=1; warn uf_urlstr("perl/camel.gif")' uf_uristr: resolving perl/camel.gif uf_uristr: gethostbyname('www.perl.bv.')...no uf_uristr: gethostbyname('www.perl.com.')...yes uf_uristr: ==> http://www.perl.com/camel.gif http://www.perl.com/camel.gif at -e line 1. ----- This is from my work computer, which is passing the test. Now, from my ubuntu box at home I get: ----- uf_uristr: resolving perl/camel.gif uf_uristr: gethostbyname('www.perl.bv.')...yes ----- So, as someone said earlier up in this conversation, sometimes 'www.perl.bv' is gettable. From work it's not, but at home it is. I'm not sure why that is the case but the solution here is that the test needs to skip if "gethostbyname('www.perl.bv')" is true. that would be a minimal fix that should stop the problem. This tests expect gethostbyname('www.perl.bv') to fail since when the author wrote it, that host did not exist. A better long term fix is to allow all these test host names be overridable via %ENV. That way if we choose bad ones we can at least override then via distroprefs while waiting for a fix. john

On Tue Jul 21 09:09:01 2009, JJNAPIORK wrote: Show quoted text
Sorry, but one more bit of information. This problem is not related to a Perl install or the way hosts or DSN works on a given machine. From my notebook computer when at work the test works properly, but when I am at home the test fails. So exact same computer, exact some perl and network setup, just a different internet provider. I guess you could say the internet providers need to fix whatever they are doing, and that would also be correct. However given the fact this causes 9/11 cpan testers failures for the current version, I'd argue strongly for the presented patch. It definitely won't hurt anything and from what I gather, the hueristics matching ability of URI is one of the more underused features in anycase.

2nd attempt and URI-1.40 uploaded to CPAN.