travis-CI failures

11 years ago by Stas Malyshev — view source

unread

Hi!

I've looked at travis-CI results for PHP (e.g.
https://travis-ci.org/php/php-src/) and I am seeing that there are two
extensions that produce fails that break the build:

Bug #39858 (Lost connection to MySQL server during query by a repeated
call stored proced) [ext/pdo_mysql/tests/bug_39858.phpt]
PDO MySQL Bug #41997 (stored procedure call returning single rowset
blocks future queries) [ext/pdo_mysql/tests/bug_41997.phpt]
MySQL PDOStatement->nextRowSet()
[ext/pdo_mysql/tests/pdo_mysql_stmt_nextrowset.phpt]
MySQL Prepared Statements and different column counts
[ext/pdo_mysql/tests/pdo_mysql_stmt_variable_columncount.phpt]
PostgreSQL notice function [ext/pgsql/tests/09notice.phpt]
Bug #32223 (weird behaviour of pg_last_notice)
[ext/pgsql/tests/80_bug32223.phpt]
Bug #32223 (weird behaviour of pg_last_notice using define)
[ext/pgsql/tests/80_bug32223b.phpt]

For pod_mysql, both bugs are listed as fixed, but they still occur and I
just checked on my own linux install, they happen exactly as on
travis-ci machine. Other ones seem to produce same "general error", so I
suspect there is a common cause behind all these. Could somebody
maintaning these take a look into them?

For PGSQL, it seems to be related to bug 32223 and notices, but I don't
have postgres install to verify it.

It would be very nice if we could get travis CI to be green (let alone
the fact that we have failures in our code base that aren't supposed to
be there).

--
Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

11 years ago by Yasuo Ohgaki — view source

unread

Hi Stas,

On Sun, Aug 11, 2013 at 12:49 PM, Stas Malyshev smalyshev@sugarcrm.comwrote:

PostgreSQL notice function [ext/pgsql/tests/09notice.phpt]
Bug #32223 (weird behaviour of pg_last_notice)
[ext/pgsql/tests/80_bug32223.phpt]
Bug #32223 (weird behaviour of pg_last_notice using define)
[ext/pgsql/tests/80_bug32223b.phpt]

For PGSQL, it seems to be related to bug 32223 and notices, but I don't
have postgres install to verify it.

It seems it's failing with PostgreSQL 8.x, I'll look into it.

Regards,

--
Yasuo Ohgaki
yohgaki@ohgaki.net

11 years ago by Yasuo Ohgaki — view source

unread

Hi Stas,

On Sun, Aug 11, 2013 at 12:49 PM, Stas Malyshev smalyshev@sugarcrm.comwrote:

PostgreSQL notice function [ext/pgsql/tests/09notice.phpt]
Bug #32223 (weird behaviour of pg_last_notice)
[ext/pgsql/tests/80_bug32223.phpt]
Bug #32223 (weird behaviour of pg_last_notice using define)
[ext/pgsql/tests/80_bug32223b.phpt]

For PGSQL, it seems to be related to bug 32223 and notices, but I don't
have postgres install to verify it.

It seems it's failing with PostgreSQL 8.x, I'll look into it.

I have PostgreSQL 8.1 though 9.3 installed on my dev system and cannot
reproduce with my PostgreSQLs and PHP 5.5 branch.

I guess Trivis-CI has some settings.

I think it can be ignored, but if there is postgresql.conf available, I'll
look
into it.

Regards,

--
Yasuo Ohgaki
yohgaki@ohgaki.net

11 years ago by Stas Malyshev — view source

unread

Hi!

I have PostgreSQL 8.1 though 9.3 installed on my dev system and cannot
reproduce with my PostgreSQLs and PHP 5.5 branch.

It happens on master, not on 5.5. Here's description of Travis CI:
http://about.travis-ci.org/docs/user/ci-environment/
http://about.travis-ci.org/docs/user/database-setup/

I guess Trivis-CI has some settings.

I think it can be ignored, but if there is postgresql.conf available,
I'll look
into it.

You mean these tests should be removed or put into XFAIL? We shouldn't
just ignore failing tests, that kind of defeats the purpose of having them.

Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

11 years ago by Yasuo Ohgaki — view source

unread

Hi Stas,

On Sun, Aug 11, 2013 at 1:43 PM, Stas Malyshev smalyshev@sugarcrm.comwrote:

I have PostgreSQL 8.1 though 9.3 installed on my dev system and cannot
reproduce with my PostgreSQLs and PHP 5.5 branch.

It happens on master, not on 5.5. Here's description of Travis CI:

pgsql module's code is the same, I think.
I run tests on master and still could not reproduce.

http://about.travis-ci.org/docs/user/ci-environment/
http://about.travis-ci.org/docs/user/database-setup/

I guess Trivis-CI has some settings.

I think it can be ignored, but if there is postgresql.conf available,
I'll look
into it.

You mean these tests should be removed or put into XFAIL? We shouldn't
just ignore failing tests, that kind of defeats the purpose of having them.

According to the doc, they have PostgreSQL 9.1.x.
However, tests for 8.x is failing which should not happen with 9.x
environment.
I guess they have non standard setup for reasons.

My PostgreSQL binary are took from git check out and built as

./configure --prefix=/some/path && make install

Only place that I've changed in postgresql.conf is port number, so that I
can
start multiple PostgreSQL at once.

I'm using

[yohgaki@dev PHP-master]$ uname -a
Linux dev 3.9.9-302.fc19.x86_64 #1 SMP Sat Jul 6 13:41:07 UTC 2013 x86_64
x86_64 x86_64 GNU/Linux

According to the doc, they are using "Ubuntu 12.04 LTS Server Edition 64
bit"
Any Ubuntu users can verify the issue?

Regards,

--
Yasuo Ohgaki
yohgaki@ohgaki.net

11 years ago by Yasuo Ohgaki — view source

unread

Hi Stas,

http://about.travis-ci.org/docs/user/ci-environment/

http://about.travis-ci.org/docs/user/database-setup/

I guess Trivis-CI has some settings.

I think it can be ignored, but if there is postgresql.conf available,
I'll look
into it.

You mean these tests should be removed or put into XFAIL? We shouldn't
just ignore failing tests, that kind of defeats the purpose of having
them.

According to the doc, they have PostgreSQL 9.1.x.
However, tests for 8.x is failing which should not happen with 9.x
environment.
I guess they have non standard setup for reasons.

Sorry I confused by the test file name.
The test works under 9.x also.

Is the diff or log files for test scripts available?
I cannot fix problem that cannot be verified.

Regards,

--
Yasuo Ohgaki
yohgaki@ohgaki.net

11 years ago by Yasuo Ohgaki — view source

unread

Sorry I confused by the test file name.
The test works under 9.x also.

I'll update test description later so that I would be confused again.

--
Yasuo Ohgaki
yohgaki@ohgaki.net

11 years ago by Stas Malyshev — view source

unread

Hi!

Is the diff or log files for test scripts available?
I cannot fix problem that cannot be verified.

Sure, right here:
https://travis-ci.org/php/php-src

Just scroll down to the end, you'll see the failed scripts. Then look
them up in the log, and you'll see the diffs.

Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

11 years ago by Stas Malyshev — view source

unread

Hi!

I think I know what the problem is. pgsql has this setting:

pgsql.ignore_notice

And the tests seem to rely on it being 0, however one of the tests does
not set it at all and another misspells it as pgsql.ignore_notices. I'll
fix it and see what happens.

--
Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

11 years ago by Stas Malyshev — view source

unread

Hi!

Hi!

I think I know what the problem is. pgsql has this setting:

pgsql.ignore_notice

And the tests seem to rely on it being 0, however one of the tests does
not set it at all and another misspells it as pgsql.ignore_notices. I'll
fix it and see what happens.

Unfortunately, this didn't help :( Still returns false for
pg_last_notice there, breaking the test. Any idea why that function can
return false in that environment?

--
Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

11 years ago by Stas Malyshev — view source

unread

Hi!

For pod_mysql, both bugs are listed as fixed, but they still occur and I
just checked on my own linux install, they happen exactly as on
travis-ci machine. Other ones seem to produce same "general error", so I
suspect there is a common cause behind all these. Could somebody
maintaning these take a look into them?

Looks like if I build with mysqlnd, it works on travis - but not on my
machine. In any case, I'll start with updating travis recipe to use
mysqlnd then.

--
Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

11 years ago by Nikita Popov — view source

unread

On Sun, Aug 11, 2013 at 5:49 AM, Stas Malyshev smalyshev@sugarcrm.comwrote:

Hi!

I've looked at travis-CI results for PHP (e.g.
https://travis-ci.org/php/php-src/) and I am seeing that there are two
extensions that produce fails that break the build:

I just fixed the last pgsql failure and Travis now has a clean build on
master.

On PHP-5.5 there is only one remaining failure, but it only occurs
sometimes, not always:

TEST 9299/12089 [ext/standard/tests/file/disk_free_space_basic.phpt]
========DIFF========
012+ Free Space Value Is Incorrect
012- Free Space Value Is Correct
========DONE========

Nikita

11 years ago by Stas Malyshev — view source

unread

Hi!

I just fixed the last pgsql failure and Travis now has a clean build on
master.

Thanks a lot! I'm afraid though there's still work to do since current
build doesn't actually seem to run all mysql tests due to some missing
settings. I'll look into it shortly.

--
Stanislav Malyshev, Software Architect
SugarCRM: http://www.sugarcrm.com/
(408)454-6900 ext. 227

travis-CI failures

You mean these tests should be removed or put into XFAIL? We shouldn't just ignore failing tests, that kind of defeats the purpose of having them.

Just scroll down to the end, you'll see the failed scripts. Then look them up in the log, and you'll see the diffs.

You mean these tests should be removed or put into XFAIL? We shouldn't
just ignore failing tests, that kind of defeats the purpose of having them.

Just scroll down to the end, you'll see the failed scripts. Then look
them up in the log, and you'll see the diffs.