Fix Solaris tests and enable CI #20709

psumbera · 2025-12-15T13:42:44Z

No description provided.

devnexen · 2025-12-15T20:43:58Z

cc @iluuu1994 at least for the CI's aspect.

devnexen · 2025-12-15T20:46:25Z

ext/pcntl/tests/pcntl_cpuaffinity.phpt

 --SKIPIF--
 <?php
 if (!function_exists("pcntl_setcpuaffinity")) die("skip pcntl_setcpuaffinity is not available");
+if (PHP_OS_FAMILY === 'Solaris') die("skip CPU affinity not supported on Solaris");


is the test failing and in which manner ? because as far as I see it is supported by solaris.

It fails like this:

Running selected tests. TEST 1/1 [ext/pcntl/tests/pcntl_cpuaffinity.phpt] ========DIFF======== 001- bool(true) 002- array(0) { 003- } 004- array(0) { 005- } 006- bool(true) 007- pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) must not be empty 008- pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) cpu id invalid value (def) 009- pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) cpu id must be between 0 and %d (%d) 010- pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) cpu id must be between 0 and %d (-1024) 011- pcntl_getcpuaffinity(): Argument #1 ($process_id) invalid process (-1024) 012- pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) value must be of type int|string, array given 001+ Warning: pcntl_setcpuaffinity(): pset_create: Cannot allocate memory in /root/php-src/ext/pcntl/tests/pcntl_cpuaffinity.php on line 3 002+ bool(false) 003+ 004+ Warning: pcntl_getcpuaffinity(): pset_create: Cannot allocate memory in /root/php-src/ext/pcntl/tests/pcntl_cpuaffinity.php on line 4 005+ 006+ Fatal error: Uncaught TypeError: array_diff(): Argument #2 must be of type array, false given in /root/php-src/ext/pcntl/tests/pcntl_cpuaffinity.php:5 007+ Stack trace: 008+ #0 /root/php-src/ext/pcntl/tests/pcntl_cpuaffinity.php(5): array_diff(Array, false) 009+ #1 {main} 010+ thrown in /root/php-src/ext/pcntl/tests/pcntl_cpuaffinity.php on line 5 ========DONE======== FAIL pcntl_getcpuaffinity() and pcntl_setcpuaffinity() [ext/pcntl/tests/pcntl_cpuaffinity.phpt]

the test might need adaptation rather than disabling it, pset_create seems to set errno to ENOMEM here. We can devise about it later..

btw do you run it under a VM ? curious also what psrset gives

btw do you run it under a VM ? curious also what psrset gives

Yes, it's VM. When I run test on real machine I get:

========DIFF======== -- pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) cpu id invalid value (def) pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) cpu id must be between 0 and %d (%d) pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) cpu id must be between 0 and %d (-1024) 011- pcntl_getcpuaffinity(): Argument #1 ($process_id) invalid process (-1024) pcntl_setcpuaffinity(): Argument #2 ($cpu_ids) value must be of type int|string, array given ========DONE======== FAIL pcntl_getcpuaffinity() and pcntl_setcpuaffinity() [ext/pcntl/tests/pcntl_cpuaffinity.phpt]

I don't see difference in psrset output between VM and real machine. Both give lines like:

processor set SYSpsrset_1 1: empty ..

Ok that should be largely enough, when I implemented this feature, it I just used a openindiana VM with only 8GB.
That might be swap space issue ; I guess you re really booting a 64 bits kernel.
That would be sad to disable this test just because of particular environment context.

Ok, there is something wrong with VM. Another one doesn't have the problem.

Not sure about this one. For now I left it disabled:

TEST 1/1 [ext/pcntl/tests/pcntl_getcpu.phpt] ========DIFF======== bool(true) 002- int(1) 002+ int(2) ========DONE======== FAIL pcntl_getcpu() [ext/pcntl/tests/pcntl_getcpu.phpt]

this one issue seems real at first glance, the test should not even launch on solaris/illumos. Is HAVE_SCHED_GETCPU commented in your main/php_config.h ?

Yes, it is (/* #undef HAVE_SCHED_GETCPU */).

devnexen · 2025-12-15T20:47:16Z

ext/opcache/shared_alloc_shm.c

 # define MIN(x, y) ((x) > (y)? (y) : (x))
 #endif

-#define SEG_ALLOC_SIZE_MAX 32*1024*1024


it seems a slightly unrelated change to me (I do not say it is right or wrong tough).

On Solaris, OPcache uses the SysV SHM allocator (shared_alloc_shm.c), which hard-limits individual segments to 32 MB (SEG_ALLOC_SIZE_MAX). With JIT enabled, OPcache reserves a default 64 MB buffer (ZEND_JIT_DEFAULT_BUFFER_SIZE) from the last shared segment during zend_shared_alloc_startup(). This makes startup fail with “Insufficient shared memory!” because the last segment can never satisfy the reservation.

sure, but what I understood from the comment is opcache on solaris can't work without this change ? If that is the case, this change needs to be a PR on its own as a bug fix (from the lowest branch it needs to apply).

I have created #20719 .

.github/actions/solaris/action.yml

Zend/tests/stack_limit/stack_limit_010.phpt

devnexen · 2025-12-16T12:42:20Z

ext/date/tests/gmstrftime_variation16.phpt

    die("skip Test is not valid for Windows");
 }
+if (PHP_OS_FAMILY === 'Solaris') {
+    die("skip Solaris uses ' 8:08:08 AM' for %r (time format differs)");


as a follow-up PR, might be worth doing solaris only versions of tests related to time formats.

devnexen · 2025-12-16T12:46:07Z

ext/date/lib/parse_date.c


 #if defined(_POSIX_TZNAME_MAX)
-# define MAX_ABBR_LEN _POSIX_TZNAME_MAX
+   /* Solaris exposes _POSIX_TZNAME_MAX = 3 unless _XPG6 is defined.


you could not know, that needs to be a PR on its own here.

This one is already in upstream:
derickr/timelib@4105b6f

The second one to this file is waiting here: derickr/timelib#169 .

I guess it might be better to update timelib in its own PR.

trusting the call to handle invalid process id via errnos. see php#20709 (comment) for rationale.

devnexen · 2025-12-19T22:57:39Z

ext/sockets/tests/socket_afpacket.phpt

 if (!defined("AF_PACKET")) {
    die('SKIP AF_PACKET not supported on this platform.');
 }
+if (!defined("ETH_P_IP")) {


oh I m surprised solaris ever supports it the same way as linux.

devnexen · 2025-12-20T11:30:44Z

ext/iconv/tests/bug37773.phpt

    die("skip UTF-8 is not supported?");
 }

+if (PHP_OS_FAMILY === 'Solaris') {


quick question, can solaris have iconv via a 3rd party package system (I guess opencsw is old now but maybe pkgsrc or similar).

we would still need to do some check in tests for such packages so there would need be some sort of skip anyway. It should not be a blocker for this though.

bukka

Nice, having Solaris in the CI would be great. It looks mostly good. Just needs some minor changes and guess we will also need to wait for #20719 .

bukka · 2025-12-23T13:40:21Z

.github/workflows/push.yml

      - name: FreeBSD
        uses: ./.github/actions/freebsd
+  SOLARIS:
+    if: github.repository == 'psumbera/php-src' || github.event_name == 'pull_request'


Suggested change

if: github.repository == 'psumbera/php-src' || github.event_name == 'pull_request'

if: github.repository == 'php/php-src' || github.event_name == 'pull_request'

bukka · 2025-12-23T13:41:45Z

.github/workflows/root.yml

      run_linux_ppc64: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 4) || matrix.branch.version[0] >= 9 }}
      run_macos_arm64: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 4) || matrix.branch.version[0] >= 9 }}
      run_freebsd_zts: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 3) || matrix.branch.version[0] >= 9 }}
+      run_solaris_zts: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 3) || matrix.branch.version[0] >= 9 }}


As we won't be backporting the test changes this is effectively ...

Suggested change

run_solaris_zts: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 3) || matrix.branch.version[0] >= 9 }}

run_solaris_zts: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 6) || matrix.branch.version[0] >= 9 }}

bukka · 2025-12-23T13:44:19Z

ext/date/lib/parse_date.c


 #if defined(_POSIX_TZNAME_MAX)
-# define MAX_ABBR_LEN _POSIX_TZNAME_MAX
+   /* Solaris exposes _POSIX_TZNAME_MAX = 3 unless _XPG6 is defined.


I guess it might be better to update timelib in its own PR.

bukka · 2025-12-23T13:56:38Z

ext/posix/tests/posix_ttyname_value_errors.phpt

+--EXPECTREGEX--
+Warning: posix_ttyname\(\): Argument #1 \(\$file_descriptor\) must be between 0 and \d+ in .+ on line \d+
+bool\(false\)
+string\(\d+\) "Bad file (descriptor|number)"


why don't you just use EXPECTF with %r(descriptor|number)%r ? That's what you could use in other tests as well

bukka · 2025-12-23T13:57:38Z

ext/standard/tests/file/007_variation1.phpt

 li"

-Notice: fwrite(): Write of 37 bytes failed with errno=9 Bad file descriptor in %s on line %d
+Notice: fwrite(): Write of 37 bytes failed with errno=9 Bad file %s in %s on line %d


Can you try with?

Suggested change

Notice: fwrite(): Write of 37 bytes failed with errno=9 Bad file %s in %s on line %d

Notice: fwrite(): Write of 37 bytes failed with errno=9 Bad file %r(descriptor|number)%r in %s on line %d

psumbera requested review from TimWolla, bukka, derickr, devnexen, dstogov, kamil-tekiela and ndossche as code owners December 15, 2025 13:42

psumbera force-pushed the github-solaris-ci branch 2 times, most recently from bbc1c07 to 9c9436d Compare December 15, 2025 14:17

devnexen reviewed Dec 15, 2025

View reviewed changes

.github/actions/solaris/action.yml Outdated Show resolved Hide resolved

psumbera force-pushed the github-solaris-ci branch from 9c9436d to 924f86d Compare December 16, 2025 09:59

devnexen reviewed Dec 16, 2025

View reviewed changes

Zend/tests/stack_limit/stack_limit_010.phpt Show resolved Hide resolved

devnexen reviewed Dec 16, 2025

View reviewed changes

psumbera mentioned this pull request Dec 16, 2025

PHP on Solaris with opcache.enable=1 and opcache.jit=1205 gets Fatal Error Insufficient shared memory! #20718

Open

devnexen added a commit to devnexen/php-src that referenced this pull request Dec 19, 2025

ext/pcntl: fix pcntl_getcpuaffinity() for solaris.

63b9ef4

trusting the call to handle invalid process id via errnos. see php#20709 (comment) for rationale.

devnexen mentioned this pull request Dec 19, 2025

ext/pcntl: fix pcntl_getcpuaffinity() for solaris. #20731

Open

kamil-tekiela removed their request for review December 19, 2025 14:48

psumbera force-pushed the github-solaris-ci branch 2 times, most recently from 92dfd8d to e0f04d9 Compare December 19, 2025 18:44

Fix Solaris tests and enable CI

106b74e

psumbera force-pushed the github-solaris-ci branch from e0f04d9 to 106b74e Compare December 19, 2025 19:49

devnexen reviewed Dec 19, 2025

View reviewed changes

devnexen reviewed Dec 20, 2025

View reviewed changes

bukka reviewed Dec 23, 2025

View reviewed changes

	if: github.repository == 'psumbera/php-src' \|\| github.event_name == 'pull_request'
	if: github.repository == 'php/php-src' \|\| github.event_name == 'pull_request'

	run_solaris_zts: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 3) \|\| matrix.branch.version[0] >= 9 }}
	run_solaris_zts: ${{ (matrix.branch.version[0] == 8 && matrix.branch.version[1] >= 6) \|\| matrix.branch.version[0] >= 9 }}

	Notice: fwrite(): Write of 37 bytes failed with errno=9 Bad file %s in %s on line %d
	Notice: fwrite(): Write of 37 bytes failed with errno=9 Bad file %r(descriptor\|number)%r in %s on line %d

Fix Solaris tests and enable CI #20709

Are you sure you want to change the base?

Fix Solaris tests and enable CI #20709

Conversation

psumbera commented Dec 15, 2025

Uh oh!

devnexen commented Dec 15, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

devnexen Dec 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

devnexen Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

devnexen Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

psumbera Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

devnexen Dec 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

devnexen Dec 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bukka left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

devnexen Dec 18, 2025 •

edited

Loading

devnexen Dec 19, 2025 •

edited

Loading

devnexen Dec 19, 2025 •

edited

Loading

psumbera Dec 19, 2025 •

edited

Loading

devnexen Dec 15, 2025 •

edited

Loading

devnexen Dec 16, 2025 •

edited

Loading