[Bug]: Test failure with 1.17.1

[remicollet]

Bug report

Number of tests :   643               563
Tests borked    :     1 (  0.2%) --------
Tests skipped   :    79 ( 12.3%) --------
Tests warned    :     0 (  0.0%) (  0.0%)
Tests failed    :    85 ( 13.2%) ( 15.1%)
Tests passed    :   478 ( 74.3%) ( 84.9%)

All of them failing because of _dd.tags.process entry in output

Ex:

========DIFF========
--
         ["type"]=>
         string(8) "new.type"
         ["meta"]=>
023-     array(1) {
023+     array(2) {
           ["_dd.base_service"]=>
           string(27) "test_special_attributes.php"
026+       ["_dd.tags.process"]=>
027+       string(142) "entrypoint.basedir:ext,entrypoint.name:test_special_attributes,entrypoint.type:script,entrypoint.workdir:datadog_trace-1.17.1,runtime.sapi:cli"
         }
         ["metrics"]=>
         array(1) {
--
========DONE========
FAIL Reserved OTel attributes that have special meaning [tests/ext/test_special_attributes.phpt] 
                                                    
========DIFF========
--
         ["type"]=>
         string(8) "new.type"
         ["meta"]=>
023-     array(1) {
023+     array(2) {
           ["_dd.base_service"]=>
           string(31) "test_special_attributes_bis.php"
026+       ["_dd.tags.process"]=>
027+       string(146) "entrypoint.basedir:ext,entrypoint.name:test_special_attributes_bis,entrypoint.type:script,entrypoint.workdir:datadog_trace-1.17.1,runtime.sapi:cli"
         }
         ["metrics"]=>
         array(1) {
--
========DONE========
FAIL Reserved OTel attributes that have special meaning [tests/ext/test_special_attributes_bis.phpt] 

Help welcome on this

PHP version

8.4.16

Tracer or profiler version

1.17.1

Installed extensions

No response

Output of phpinfo()

No response

Upgrading from

No response

[bwoebi]

Ah I see, we recently added a feature datadog.experimental_propagate_process_tags_enabled=1 (i.e. enabled by default).
When we implemented this we took the shortcut of simply running our tests with the INI globally disabled, but that's really not how it should be.

I.e., for this time if you set -d datadog.experimental_propagate_process_tags_enabled=0 to the test runner, it should work.

Thanks for calling this out. We'll fix this properly.

[remicollet]

Better for test.... but now some segfaults

=====================================================================
Number of tests :   643               562
Tests borked    :     2 (  0.3%) --------
Tests skipped   :    79 ( 12.3%) --------
Tests warned    :     0 (  0.0%) (  0.0%)
Tests failed    :     0 (  0.0%) (  0.0%)
Tests passed    :   562 ( 87.4%) (100.0%)
---------------------------------------------------------------------
Time taken      : 20.674 seconds
=====================================================================

=====================================================================
BORKED TEST SUMMARY
---------------------------------------------------------------------
Termsig=0 [/builddir/build/BUILD/php-pecl-datadog-trace-1.17.1-build/php-pecl-datadog-trace-1.17.1/datadog_trace-1.17.1/tests/ext/background-sender/sidecar_thread_mode_permissions.phpt]
Termsig=0 [/builddir/build/BUILD/php-pecl-datadog-trace-1.17.1-build/php-pecl-datadog-trace-1.17.1/datadog_trace-1.17.1/tests/ext/pcntl/pcntl_fork_thread_mode_orphan.phpt]
=====================================================================

[bwoebi]

Note: It won't be fixed yet for 1.18.0 (we do that one quicker than I expected. I'm still going to look into this.).

[bwoebi]

I have no idea why these two tests crash - they don't for us.
The datadog.experimental_propagate_process_tags_enabled=0 no longer needs to be set manually starting next release.

[bwoebi]

@remicollet 1.19 no longer requires that INI, but I've never been able to reproduce those borked tests.

[remicollet]

Still have (1.19.0) the 2 bork tests

BORK Termsig=0 [tests/ext/background-sender/sidecar_thread_mode_permissions.phpt] reason: invalid output from SKIPIF
BORK Termsig=0 [tests/ext/pcntl/pcntl_fork_thread_mode_orphan.phpt] reason: invalid output from SKIPIF

[bwoebi]

@remicollet Yeah, it would be nice if you could show me the output you see in your SKIPIFs - I have no idea what goes wrong.

[remicollet]

@remicollet Yeah, it would be nice if you could show me the output you see in your SKIPIFs - I have no idea what goes wrong.

Will try to get more info... but for now my builder is busy with building 1.19.0 (without tests)... for >5h...
So don't expect anything quickly