Hi Remi!
I see that the test sapi/fpm/tests/015.phpt is randomly intermittently
failing all the time in CI, making it harder to evaluate fixes. The
cause of the failure is additional one or two messages like this:
ERROR: unable to read what child say: Bad file descriptor (9)
Could you (or somebody else who knows FPM better than me) look into it
and make it work or XFAIL it or disable it if it's impossible?
Thanks,
Stas Malyshev
smalyshev@gmail.com
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Le 01/12/2014 04:37, Stanislav Malyshev a écrit :
Hi Remi!
I see that the test sapi/fpm/tests/015.phpt is randomly
intermittently failing all the time in CI, making it harder to
evaluate fixes. The cause of the failure is additional one or two
messages like this:ERROR: unable to read what child say: Bad file descriptor (9)
Could you (or somebody else who knows FPM better than me) look into
it and make it work or XFAIL it or disable it if it's impossible?
015 is the only test which rely on "catch_workers_output = yes"
So this feature seems to have issues :(
I think we should be able to write directly in the master error_log
without having to go through via this master-child-pipe solution.
I already have a patch and run some tests on this.
I will look at this ASAP (probably using XFAIL for now).
Notice: I think some downstream use php_admin_value[error_log]
which create a "per-pool" file, but owned by the non privileged child
user (not perfect for security, as the dir must be widely open).
With default config... (no catch_workers_output and no
php_admin_value) child messages are simply lost in /dev/null :(
Remi.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
iEYEARECAAYFAlR8BhIACgkQYUppBSnxahgNVgCfciNCGIvbvLqU13vA39wm3FHl
SoAAniqUOWBNMavbltgh0E3vi0FuKAqb
=rU2D
-----END PGP SIGNATURE
On Mon, Dec 1, 2014 at 4:37 AM, Stanislav Malyshev smalyshev@gmail.com
wrote:
Hi Remi!
I see that the test sapi/fpm/tests/015.phpt is randomly intermittently
failing all the time in CI, making it harder to evaluate fixes. The
cause of the failure is additional one or two messages like this:ERROR: unable to read what child say: Bad file descriptor (9)
Could you (or somebody else who knows FPM better than me) look into it
and make it work or XFAIL it or disable it if it's impossible?Thanks,
Stas Malyshev
smalyshev@gmail.com--
hi Stas,
where do you see this test failing?
on travis, or somewhere else?
--
Ferenc Kovács
@Tyr43l - http://tyrael.hu
Hi!
where do you see this test failing?
on travis, or somewhere else?
Yes, on Travis. Fails almost every other time.
--
Stas Malyshev
smalyshev@gmail.com
On Tue, Dec 2, 2014 at 8:26 PM, Stanislav Malyshev smalyshev@gmail.com
wrote:
Hi!
where do you see this test failing?
on travis, or somewhere else?Yes, on Travis. Fails almost every other time.
--
Stas Malyshev
smalyshev@gmail.com
interesting, I've checked out https://travis-ci.org/php/php-src/builds
before sending my mail, and looked into a couple of failed builds, but
couldn't find anything with this particular test.
--
Ferenc Kovács
@Tyr43l - http://tyrael.hu
interesting, I've checked out https://travis-ci.org/php/php-src/builds
before sending my mail, and looked into a couple of failed builds, but
couldn't find anything with this particular test.
As far as I can see, Remi already marked it as XFAIL.
2014.12.02. 20:57 ezt írta ("Maciej Sobaczewski" sobak@php.net):
interesting, I've checked out https://travis-ci.org/php/php-src/builds
before sending my mail, and looked into a couple of failed builds, but
couldn't find anything with this particular test.As far as I can see, Remi already marked it as XFAIL.
Yes. I was looking before it was marked though.
But now I was able to reproduce albeit like once in a hundred times.