Everything

php

"forced" ini settings in CLI don't work anymore..

php.internals

Jani Taskinen

23 years ago

Apparently the changes you made to cleanup the ini settings
forcing in CLI do not work as expected:

<?php

var_dump(ini_get('register_argc_argv'));
var_dump(ini_get('html_errors'));
var_dump(ini_get('display_errors'));
var_dump(ini_get('implicit_flush'));
var_dump(ini_get('max_execution_time'));

?>

Results with CLI build before the changes were merged:

string(1) "1"
string(1) "0"
string(1) "1"
string(1) "1"
string(1) "0"

Results with latest CVS checkout (PHP_4_3 branch):

string(1) "1"
string(1) "0"
string(1) "1"
string(0) ""
string(2) "30"

When '-n' is used (to not use any php.ini) both builds work fine.
So, are these not supposed to be forced defaults anymore for CLI?

--Jani

(Marcus Börger)

23 years ago

At 05:03 16.04.2003, Jani Taskinen wrote:

>
>     When '-n' is used (to not use any php.ini) both builds work fine.
>     So, are these not supposed to be forced defaults anymore for CLI?
>

They only overwrite the settings for -n now. The key idea was
that the user should be able to change the values by all three
mechanisms: "-d" option, php-cli.ini and ini_set(). If there
were any ini setting that really needed to overwrite the ini
file always we could do that, too (but i don't see a need for it).

marcus

Jani Taskinen

23 years ago

On Wed, 16 Apr 2003, Marcus Börger wrote:

>At 05:03 16.04.2003, Jani Taskinen wrote:
>>
>>     When '-n' is used (to not use any php.ini) both builds work fine.
>>     So, are these not supposed to be forced defaults anymore for CLI?
>>
>
>
>They only overwrite the settings for -n now. The key idea was
>that the user should be able to change the values by all three
>mechanisms: "-d" option, php-cli.ini and ini_set(). If there
>were any ini setting that really needed to overwrite the ini
>file always we could do that, too (but i don't see a need for it).

You effectively broke BC with this change.

See the ext/pcntl/tests/001.phpt which now fails because of this.
It assumes implicit_flush is always 'On' as it only runs in CLI.
I would guess some other people have made the same assumption..

The settings were "hardcoded" for CLI for a reason, so please
make these behave like they did before.

--Jani

(Marcus Börger)

23 years ago

At 01:33 18.04.2003, Jani Taskinen wrote:

>    You effectively broke BC with this change.
>
>     See the ext/pcntl/tests/001.phpt which now fails because of this.
>     It assumes implicit_flush is always 'On' as it only runs in CLI.
>     I would guess some other people have made the same assumption..
>
>     The settings were "hardcoded" for CLI for a reason, so please
>     make these behave like they did before.

implicit_flush *may* be the exception which *could* still be done.
However the change was on the TODO of CLI so i would favor
fixing the tests by use of --INI-- section.

marcus

(Marcus Börger)

23 years ago

At 01:33 18.04.2003, Jani Taskinen wrote:

>On Wed, 16 Apr 2003, Marcus Börger wrote:
>
> >At 05:03 16.04.2003, Jani Taskinen wrote:
> >>
> >>     When '-n' is used (to not use any php.ini) both builds work fine.
> >>     So, are these not supposed to be forced defaults anymore for CLI?
> >>
> >
> >
> >They only overwrite the settings for -n now. The key idea was
> >that the user should be able to change the values by all three
> >mechanisms: "-d" option, php-cli.ini and ini_set(). If there
> >were any ini setting that really needed to overwrite the ini
> >file always we could do that, too (but i don't see a need for it).
>
>     You effectively broke BC with this change.
>
>     See the ext/pcntl/tests/001.phpt which now fails because of this.
>     It assumes implicit_flush is always 'On' as it only runs in CLI.
>     I would guess some other people have made the same assumption..
>
>     The settings were "hardcoded" for CLI for a reason, so please
>     make these behave like they did before.

It is more likely that you use an ini file that is not appropriate, means
we have to
add more ini defaults in run-tests.php. For me everything is just fine
whether that
implicit_flush is On or Off doesn't make a difference for me.

marcus

Jani Taskinen

23 years ago

On Sat, 19 Apr 2003, Marcus Börger wrote:

>At 01:33 18.04.2003, Jani Taskinen wrote:
>>On Wed, 16 Apr 2003, Marcus Börger wrote:
>>
>> >At 05:03 16.04.2003, Jani Taskinen wrote:
>> >>
>> >>     When '-n' is used (to not use any php.ini) both builds work fine.
>> >>     So, are these not supposed to be forced defaults anymore for CLI?
>> >>
>> >
>> >
>> >They only overwrite the settings for -n now. The key idea was
>> >that the user should be able to change the values by all three
>> >mechanisms: "-d" option, php-cli.ini and ini_set(). If there
>> >were any ini setting that really needed to overwrite the ini
>> >file always we could do that, too (but i don't see a need for it).
>>
>>     You effectively broke BC with this change.
>>
>>     See the ext/pcntl/tests/001.phpt which now fails because of this.
>>     It assumes implicit_flush is always 'On' as it only runs in CLI.
>>     I would guess some other people have made the same assumption..
>>
>>     The settings were "hardcoded" for CLI for a reason, so please
>>     make these behave like they did before.
>
>
>It is more likely that you use an ini file that is not appropriate, means
>we have to add more ini defaults in run-tests.php. For me everything is
>just fine whether that implicit_flush is On or Off doesn't make a difference for me.

What part in "you broke BC" you don't understand?
It worked BEFORE your patch, it doesn't work AFTER your patch.
(I had the same php.ini for both tests)

--Jani