Please fork https://github.com/mvorisek/php-src/tree/fix_oci8_mem_leak branch, it contains php-src test case.
This PR might decrease the CI quality. Imagine a test with failure 5% failure rate (or 10k tests with 5ppm failure rate). If they are retried twice, instead of such random failure being detected in 1 in 20 runs, in will be detected in 1 in 400 tests. I know it can display WARN log, but not many people read it, one example is #10838 which noone noticed/reported for 1/2 year.
My question is if "timeout CI failures" like https://github.com/php/php-src/actions/runs/4502125622/jobs/7923508402#step:13:132 can be fixed better. From my experience, they can, like using nice
for the (web)server to have higher CPU priority (than the client).
@bwoebi yes - yes, I know, but I say in reality they can be uncaught twice, once in non-shutdown and once in shutdown
when set_exception_handler
set from non-shutdown can handle shutdown exception, it should handle it even if it already handled one in non-shutdown
Excluded custom props for now, but it's wrong. Not sure how to fix,
Can you link the fix PR here?
I am not entirely sure what the expected result should be for JSON, I would probably expect the same JSON format as in standard JSON/JS.
An exact case might be expected by some user code and some transformations like camel<>snake case, I would prefer this to be fixed,
I personally would expect the handler to handle an uncaught exception in shutdown as well even if it already handled one exception in non-shutdown already, it would be more consistent with set_error_handler
which handle multiple warnings as well. To assert the behaviour, feel free to integrate full test case from #10695.
@fisker there are more review now, would it be possible to merge this PR so it can be part of the next release?
About your 2nd point - why? please point to vue code, as ID can be a string like UUID or any other string.
@Wirone please see https://github.com/phpstan/phpdoc-parser/pull/143, escaped phpdoc string is valid and must be supported, it is obvious the original \'[^\']+?\'
regex cannot handle escaped quotes in quotes.