Randy Kobes wrote:I didn't see it like that: This test fails intermittently, and doesn't require any special combination of prior tests to make it fail - it even fails when run on its own. The other filter test that we're looking at (both_str_con_add.t) only fails when run after certain other tests (t/api + t/apr), and does so every time under those circumstances, but it never fails when run on its own.
On Thu, 18 Sep 2003, Stas Bekman wrote:
Steve Hay wrote:
[ .. ]
Damn. I've just discovered that this isn't actually fixed -- sometimes
it works, sometimes not.
I just ran "perl t/TEST t/filter/in_str_consume.t" six times and got these results:
FAIL PASS FAIL PASS PASS FAIL
What's the failure? All the same? A silent die? This is only with APACHE_TEST_PRETEND_NO_LWP=1, right?
I get the same failure, without APACHE_TEST_PREPEND_NO_LWP,
but not consistently (occasionally it'll pass). When it
fails the test returns nothing. I wasn't too worried about
it at the time, as all the filter tests have problems when
run from 'nmake test', and so this may ultimately get
resolved when that problem is fixed (maybe).
They sound like different problems to me, but who knows...
Man, people are going to have so much fun with this random thing when 5.8.1 hits the streets. If you run your tests with env PERL_HASH_SEED_DEBUG=1 you will be able to re-use the HASH_SEED it reported when you had a failure to reliable reproduce the problem if this has anything to do with the hash randomness. it's possible that it's not the issue, but I'd try that first. To tell perl to use a specific seed just run the test with:
env PERL_HASH_SEED=9879434 t/TEST ... (and the seed comes from the mod_perl report
it error_log)
It's nothing to do with hash seed randomisation: I grabbed the hash seed from a "successful" run from the error_log and ran a few times with that set in the env and got some failures. Conversely, running with a hash seed from a "failed" run, I got some successes.
You say: without APACHE_TEST_PREPEND_NO_LWP, so you do have LWP, right and it doesn't use Apache::TestClient, so the problem is elsewhere and not in the broken pipe issue that is now fixed.
I also have LWP. I've tried running this test with and without APACHE_TEST_PREPEND_NO_LWP set and get a mix of successes/failures either way.
also do you have this problem when running standalone, or with other tests?
This one fails when run on its own (perl t/TEST t/filter/in_str_consume.t), or when run with other tests.
- Steve
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
