Can we get a backport of this for 3.10 too?

Maybe a better do not add $metric when it too big, because adding "half data" will show incorrect "timing", that may mislead.
Or I missed something?

@zero-24 we do not have it in 3.10, it was a new feature in 4.1 ?

Ha ok no backport needed in this case :D

@laoneo Does this PR solve issue #37080 so the issue can be closed?

@laoneo Does this PR solve issue #37080 so the issue can be closed?

At the end of the day it should solve it

Maybe a better do not add $metric when it too big, because adding "half data" will show incorrect "timing", that may mislead. Or I missed something?

I was thinking the same, perhaps we should then return one entry with a message too much data.

In reference to #37080 if I lower the strlen value to say 3600 IT WORKS. With the current PR value (4000) I still have problems.

@laoneo do you think it makes sense to make that a option within the plugin with a default value for 4k (Yes I also hate new options)? So on sites such as @mikeazores 's he can lower the value but on other sites we can use the 4k / 8k etc.?

I was thinking the same, perhaps we should then return one entry with a message too much data.

Or maybe do only 2 point timing: between start-end (when too big).
Sadly we cannot easily exclude timings for modules.

@laoneo do you think it makes sense to make that a option within the plugin with a default value for 4k (Yes I also hate new options)? So on sites such as @mikeazores 's he can lower the value but on other sites we can use the 4k / 8k etc.?

For me it should work properly out of the box and it is not that critical, that it must always return a value.

I was thinking the same, perhaps we should then return one entry with a message too much data.

Or maybe do only 2 point timing: between start-end (when too big). Sadly we cannot easily exclude timings for modules.

We can, and deliver module rendering as one metric when there is a string like mod_ in the profile step.

Sorry, 4000 limit is too large, I had to make it 3500 to get it to work on my server.
See #37090

The server-wide Nginx settings for my server are 'out of the box'.

Can you making it configurable on the debug plugin parameters?
Default to 2000?
Warning about needing to change nginx settings if you increase it?
Suggest those nginx settings I grabbed earlier as examples?

@BrainforgeUK probably the better approach would be to collect the module ones as one entry. Having an option for such a small feature would be overhead.

The last commit adds a new behavior where the modules are collected together and are printed as one entry. @Fedik what do you think?

It looks then like:

Looks good to me.

Additionally still better to keep your previous approach with limit check, just to be safe, around 3K.
Because we cannot guarantee size when any extensions will add own Profiling data.

Added a limit again. Can you guys test it now so we can ship it with the next patch release?

Can you Plesse add 2 things?

1. make the limit a parameter
2. Add a line that to the metric that the limit got hit.

So it's possible to know that something is missing and that you can adjust it depending on your needs.

Adding a new parameter for such a small feature makes no sense and the majority will not even understand for what it is.

The parameter is in the debug plugin, the person configuring the plugin should know what the web does but OK, it's not important I guess. But the information that the data is truncated is important from my point of view.

How would you render that message?

How would you render that message?

Kind of last row in metrics

$metrics .= sprintf('%s;dur=%f;desc="%s", ', $index . 'Rest of data is gone forever', 'infinity', '');

?

Last commit shows now a truncated message:

If you want to test it, change the code if (strlen($metrics) > 3000) to if (strlen($metrics) > 1).

I have tested this item ✅ successfully on 9767226

@BrainforgeUK and @mikeazores can you test it again as well?

I have tested this item ✅ successfully on 9767226

RTC

Back to pending due to new changes.

Back to pending due to new changes.

We had the same problem with access when a none admin get logged in on the back end. The recent changes are grouping also access.

I have tested this item ✅ successfully on 105dc3d

I have tested this item ✅ successfully on 105dc3d

RTC

Can't we just put the array in a variable, count and slice out the items which are too much from the middle (https://www.php.net/manual/en/function.array-splice.php) so we have a start/end time?

You just have a time of a task, there is no start/end time. This is not a whaterfall chart.

I'm setting this one as a release blocker as it affects a lot of sites and should definitely be fixed and finally shipped with 4.1.1.

@laoneo Well, the last comment was from @bembelimen , so it would be good to know if he is ok with your reply to it,

He didn't reply so my assumption was that he did understand my explanation. I mean the pr is so correct, if he wants to make some adjustments, then he can do that in a followup pr.

Sure.

Thx