GitHub Actions CI #6

NattyNarwhal · 2022-02-09T20:10:42Z

This:

Builds the driver on all supported PHP versions for Windows (7.4, 8.0, 8.1)
Builds the driver on all supported PHP versions for Ubuntu (7.3, 7.4, 8.0, 8.1), starts Db2 for Linux in a Docker container, and runs the test suite on it
Makes any changes needed to make the test suite pass

kadler · 2022-02-09T21:38:25Z

.github/workflows/ci.yml

+    runs-on: ubuntu-latest
+    steps:
+      - name: Install DB2 library
+        # XXX: Can this be optimized to not constantly redownload? Or be put in a better place?


https://github.com/actions/cache

Definitely would be helpful to cache, I think. I know the PRs for python-ibmdb were failing for a while due to flakiness in the IBM download site.

NattyNarwhal · 2022-02-10T18:26:27Z

Beaten into better shape, though the PHP 8.0 failure is bizarre; it's as if it was trying to link it as a normal executable instead of a library?

kadler · 2022-02-10T20:41:22Z

Beaten into better shape, though the PHP 8.0 failure is bizarre; it's as if it was trying to link it as a normal executable instead of a library?

make V=1?

kadler · 2022-02-10T23:15:22Z

Seems to be some something weird with libtool. It's calling libtool --mode=compile cc which seems to be not passing the -c flag.

kadler · 2022-02-11T16:37:13Z

php_pdo_ibm.h

 #ifdef PASE /* i5/OS ease of use turn off/on */
+ZEND_BEGIN_MODULE_GLOBALS(pdo_ibm)


Another option would be to define a "uses globals" flag instead of hardcoding to PASE in case other platforms need it in the future, eg.

#if defined(PASE) #define PDO_IBM_USES_GLOBALS #endif #ifdef PDO_IBM_USES_GLOBALS ZEND_BEGIN_MODULE_GLOBALS(pdo_ibm) ... #endif Of course, that might be needless future-proofing.

Not a bad idea, but considering ibm_db2 went a long while without any non-PASE globals, might also be needless, yeah.

.github/workflows/ci.yml

The PDO extensions bundled with PHP moved to this style (well, they don't include the legacy check; this mirrors how it was in 7.3) Pushed to CI branch to see how it affects it.

This only builds the driver on Ubuntu. macOS and Windows, as well as starting the LUW Docker container are TODO.

…nt .lib

MSVC doesn't allow it, so should unbreak Windows builds. Made to CI branch, but can cherry pick back to main.

Christoph pointed out this was wrong with the new macro; very subtle!

In spirit to what PHP does with MSSQL.

Add additional environment variables for the DSN INI et al

the "automatic storage yes" is something I remembered last time I had to deal with LUW. Try to remove tty/interactive flags from docker exec Try to wait for Db2 to finish up setup (not ideal) We don't need to create the database, the Docker container made one all along Actually check if setup is complete

It's hard to predict what we'll be running on.

Do fwrite instead.

The 1/2 keys will always be empty, whereas before they had some cruft about non-existent files. In theory we could force it to be like how it was, but no other drivers do so.

I'm not sure how the original version of the test ever made sense when I was reading it. The behaviour it expects seemed counter-intuitive with how it was expecting casting to happen. Bizarrely, the IBM i version does match the behaviour of modern PHP and Db2. So, let's just put that in its place.

Not sure how to get this working for now, but we can revisit it.

It seems on PHP 7.x, the order in which they're inserted matters, at least for `fvt_025_ColumnMetaData`. Tested on PHP 7.4/LUW.

On PHP 8.1, the skip cache seemingly collates all the reused skip sections into a single file. However, the skip check relies on using the name of the test to determine if it should be skipped based on platform. This is why the skip blocks that *aren't* the same will trigger reliably. This will slow it down a smidge, but work reliably until the skip checks are refactored.

One of the LOB tests exhibited this. I don't know if NULL is the best thing to return, but it can be considered.

kadler reviewed Feb 9, 2022

View reviewed changes

NattyNarwhal force-pushed the ci branch from 40eacd1 to 0e5d8eb Compare February 10, 2022 18:03

NattyNarwhal force-pushed the ci branch from c35d420 to 474bf9f Compare February 11, 2022 15:45

kadler reviewed Feb 11, 2022

View reviewed changes

NattyNarwhal mentioned this pull request Feb 11, 2022

GitHub Actions CI php/pecl-database-ibm_db2#21

Merged

kadler reviewed Mar 2, 2022

View reviewed changes

.github/workflows/ci.yml Outdated Show resolved Hide resolved

NattyNarwhal changed the title ~~WIP: GitHub Actions CI~~ GitHub Actions CI Mar 18, 2022

NattyNarwhal added 20 commits March 21, 2022 16:42

Use PHP 7.x style PDO include check

f1b5253

The PDO extensions bundled with PHP moved to this style (well, they don't include the legacy check; this mirrors how it was in 7.3) Pushed to CI branch to see how it affects it.

Initial GitHub Actions CI

c278699

This only builds the driver on Ubuntu. macOS and Windows, as well as starting the LUW Docker container are TODO.

Basic Windows support

addd471

Only run CI on our main branch and PRs targeting it

2374199

Attempt to work around that amd64 version of clidriver uses a differe…

f597c06

…nt .lib

Break up Ubuntu steps and use verbose make

baf74e0

Don't have an empty globals on non-i

3a1b190

MSVC doesn't allow it, so should unbreak Windows builds. Made to CI branch, but can cherry pick back to main.

Fix PDO include path when using modern m4 macro

97d68f9

Christoph pointed out this was wrong with the new macro; very subtle!

Attempt to run the Db2 container

359efea

In spirit to what PHP does with MSSQL.

Attempt to fix loading PDO when running tests

a6bf66e

Fix broken variable in PASE part of tests

a3d5671

Set PDO test variables as needed

0d39d0e

Add additional environment variables for the DSN INI et al

Make options appendable in connection string

9d8bc0b

Start test adaptations for PHP 8 and modern Db2

2d0975e

The workstation name is the hostname

959d7cd

It's hard to predict what we'll be running on.

We can't copy from a string to a stream

213540a

Do fwrite instead.

PDO changed error reporting message keys

b13bee7

The 1/2 keys will always be empty, whereas before they had some cruft about non-existent files. In theory we could force it to be like how it was, but no other drivers do so.

Db2 has changed this message nowadays

deb1cda

NattyNarwhal added 8 commits March 21, 2022 16:46

Attempt to cache the Docker container

8397c44

Trusted user context requires setup

a89356a

Not sure how to get this working for now, but we can revisit it.

7.3 has been dropped from Windows support

06dd54f

Flip order of adding native_type/table

ee10e48

It seems on PHP 7.x, the order in which they're inserted matters, at least for `fvt_025_ColumnMetaData`. Tested on PHP 7.4/LUW.

Fix crash if stream for LOB is null on 8.1

f79f0dd

One of the LOB tests exhibited this. I don't know if NULL is the best thing to return, but it can be considered.

Suppress 8.1 deprecation message for NULL in PDO arg

617f25d

Suppress 8.1 deprecation msg, fix indentation

0c75465

NattyNarwhal force-pushed the ci branch from 2849204 to 0c75465 Compare March 21, 2022 19:47

NattyNarwhal merged commit e827e37 into master Mar 21, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

GitHub Actions CI #6

GitHub Actions CI #6

Uh oh!

NattyNarwhal commented Feb 9, 2022 •

edited

Loading

Uh oh!

kadler Feb 9, 2022

Uh oh!

kadler Feb 9, 2022

Uh oh!

NattyNarwhal commented Feb 10, 2022

Uh oh!

kadler commented Feb 10, 2022

Uh oh!

kadler commented Feb 10, 2022

Uh oh!

kadler Feb 11, 2022

Uh oh!

NattyNarwhal Feb 11, 2022

Uh oh!

Uh oh!

Uh oh!

		#ifdef PASE /* i5/OS ease of use turn off/on */
		ZEND_BEGIN_MODULE_GLOBALS(pdo_ibm)

GitHub Actions CI #6

GitHub Actions CI #6

Uh oh!

Conversation

NattyNarwhal commented Feb 9, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kadler Feb 9, 2022

Choose a reason for hiding this comment

Uh oh!

kadler Feb 9, 2022

Choose a reason for hiding this comment

Uh oh!

NattyNarwhal commented Feb 10, 2022

Uh oh!

kadler commented Feb 10, 2022

Uh oh!

kadler commented Feb 10, 2022

Uh oh!

kadler Feb 11, 2022

Choose a reason for hiding this comment

Uh oh!

NattyNarwhal Feb 11, 2022

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

NattyNarwhal commented Feb 9, 2022 •

edited

Loading