I'm prepping a new version of Test.Simple and related modules, and would greatly appreciate it if everyone would check it out from SVN and run the tests in their favorite browsers. You can do a quick online run here:
> And all tests pass in those browsers. How about for you? IE users > especially wanted.
Seems to get caught in a timeout loop in IE6 (SP2 on XP) in the builder_death.js test within the Test.Harness.Browser.prototype.runTests method (the scoped wait function specifically).
> Seems to get caught in a timeout loop in IE6 (SP2 on XP) in the > builder_death.js test within > the Test.Harness.Browser.prototype.runTests method (the scoped wait > function specifically).
Marvin, can you have a look at this? I changed that test to call _finish() instead of _ending() to eliminate an error in Firefox, but maybe that's not the right thing to do for IE?
On Apr 1, 2008, at 9:47 AM, David E. Wheeler wrote:
>> Seems to get caught in a timeout loop in IE6 (SP2 on XP) in the >> builder_death.js test within >> the Test.Harness.Browser.prototype.runTests method (the scoped wait >> function specifically).
> Marvin, can you have a look at this? I changed that test to call > _finish() instead of _ending() to eliminate an error in Firefox, but > maybe that's not the right thing to do for IE?
I can confirm that on IE6 Win2k, calling _ending() completes the test cleanly while calling _finish() hangs.
What sort of a problem have you run into with FireFox? With the original version (using _ending()), the builder_death.js test passes on...
Win2k SP4: IE 6.0.2800.1106 FF 2.0.0.7 Vista: IE 7.0.6000.16609 FF 2.0.0.11 OS X 10.5: Safari 3.1 (5525.13) FF 2.0.0.13
Background on builder_death.js here: http://xrl.us/biqyj (Link to groups.google.com)
> I can confirm that on IE6 Win2k, calling _ending() completes the test > cleanly while calling _finish() hangs.
Cute.
> What sort of a problem have you run into with FireFox?
I get:
Error: Test is not defined
All tests pass, so I'm not sure what's up with that. Hrm…looking more closely, if I skip the builder_death test, I don't get the error. But if I put it in, the error happens after curr_test runs, not at the end of builder_death. There must be some side-effect.
> With the > original version (using _ending()), the builder_death.js test passes > on...
On Apr 1, 2008, at 2:46 PM, David E. Wheeler wrote:
> Are you certain that the call to _ending() is necessary? See this bit > in Test.Harness.Browser:
> // IE 6 SP 2 doesn't seem to run the onload() event, so we force > the > // issue. > Test.Builder._finish(Test);
> So the harness should finish the job, no?
Yes, leaving it out works -- in all 6 browsers, and with no errors reported in the FF console.
Actually, I note that the last version of the patch I submitted, archived at <http://xrl.us/biq9g>, didn't include the force. Forcing was only needed in the earlier versions where the the test was packaged as an html file, builder_death.html, rather than as builder_death.js.
In other news, async.js dies after the first test in Vista IE7.
Unfortunately it's hard to see that it's failing when you're running the suite via the Harness, because the Harness doesn't notice that the test died and that not enough tests ran. Harness still reports "All tests successful" and just indicates a lower number than it should. The false negative arises because of the long-standing, hard to fix bug discussed at <http://xrl.us/birbc>.
>> // IE 6 SP 2 doesn't seem to run the onload() event, so we force >> the >> // issue. >> Test.Builder._finish(Test);
>> So the harness should finish the job, no?
> Yes, leaving it out works -- in all 6 browsers, and with no errors > reported in the FF console.
Oh, good. Removing it, then. Thanks!
> Actually, I note that the last version of the patch I submitted, > archived at <http://xrl.us/biq9g>, didn't include the force. Forcing > was only needed in the earlier versions where the the test was > packaged as an html file, builder_death.html, rather than as > builder_death.js.
Yeah, builder_death.html is where I copied the code from, somewhere in the archives, since it wasn't in the new svn. Good spot, thanks.
> In other news, async.js dies after the first test in Vista IE7.
Bleh.
> Unfortunately it's hard to see that it's failing when you're running > the suite via the Harness, because the Harness doesn't notice that the > test died and that not enough tests ran. Harness still reports "All > tests successful" and just indicates a lower number than it should. > The false negative arises because of the long-standing, hard to fix > bug discussed at <http://xrl.us/birbc>.
Yes, I thought TestResult was supposed to fix that. But I guess it was mainly just a refactoring, eh?
So we nee some way to get that data to the harness from Test.Builder.
According to my local copy, those files *were* committed:
/Volumes/Bruin/Users/marvin/projects/ $ cp -R Test.Simple/ Test.Simple.old /Volumes/Bruin/Users/marvin/projects/ $ cd Test.Simple.old/ /Volumes/Bruin/Users/marvin/projects/Test.Simple.old/ $ svn stat M tests/instance.js M tests/harness.js M doc/pod/Test/Harness/Browser.pod /Volumes/Bruin/Users/marvin/projects/Test.Simple.old/ $ svn revert -R . Reverted 'tests/instance.js' Reverted 'tests/harness.js' Reverted 'doc/pod/Test/Harness/Browser.pod' /Volumes/Bruin/Users/marvin/projects/Test.Simple.old/ $ rm tests/ builder_death.js /Volumes/Bruin/Users/marvin/projects/Test.Simple.old/ $ svn stat ! tests/builder_death.js /Volumes/Bruin/Users/marvin/projects/Test.Simple.old/ $
Something got f***ed up somewhere else -- my guess is that whoever re- imported the svn repository did so from an out-of-date local copy. Whoever that is should check the revision number of their copy and see what else may have gotten lost.
PS: Google Groups sux because it will have auto-rewrapped many of the lines above.
> Something got f***ed up somewhere else -- my guess is that whoever re- > imported the svn repository did so from an out-of-date local copy.
Yep. Did you commit those files in a commit separate from when you modified the other files that referred to them? For example, all the code had been updated to use TestResult, it's just that the file wasn't there.
On Apr 1, 2008, at 4:28 PM, David E. Wheeler wrote:
>> Unfortunately it's hard to see that it's failing when you're running >> the suite via the Harness, because the Harness doesn't notice that >> the >> test died and that not enough tests ran. Harness still reports "All >> tests successful" and just indicates a lower number than it should. >> The false negative arises because of the long-standing, hard to fix >> bug discussed at <http://xrl.us/birbc>.
> Yes, I thought TestResult was supposed to fix that. But I guess it was > mainly just a refactoring, eh?
It was a start. The idea was to beef up the OO discipline level a bit, and break a chunk out of the gnarly center. It's pretty hard to follow what's going on in there -- and the original Perl Test::Simple is harder still. It's going to take more refactoring yet to get to the point where internal test result tracking makes sense to an ordinary human brain.
I think you should release anyway. async.js fails in IE7 Vista in the current release (0.21), so you're not losing anything AFICT.
On Apr 1, 2008, at 4:40 PM, David E. Wheeler wrote:
> On Apr 1, 2008, at 16:36, Marvin Humphrey wrote:
>> Something got f***ed up somewhere else -- my guess is that whoever >> re- >> imported the svn repository did so from an out-of-date local copy.
> Yep. Did you commit those files in a commit separate from when you > modified the other files that referred to them? For example, all the > code had been updated to use TestResult, it's just that the file > wasn't there.
It's hard to know. Since the old repository is gone, I can't perform an "svn log" or an "svn diff" and AFAIK the commits weren't being archived anywhere at that point. From spelunking the .svn/ files, it looks like TestResult.js got added during 518.
I performed a "diff -ru" from my old local copy against the new repos version (attached). Allowing for the other changes since mine, it looks like we're square.
