XF 2.3 Some general tips for upgrading to Xenforo 2.3

[Mbx]

I would not try to use it with PHP 8.4 as it produces Zend memory errors that point to some problems (would be too extensive to go into detail here, several issues). Another problem if you update and use the new image optimization -- the WebP settings are somewhat low (seem to be calculated), and lead to a drop in quality compared to 85% JPEG. WebP should be manually set to 90% quality right in beginning. Why? Because instead of saving the WebP files separately (which would been better, like for checking support and then choosing between JPEG and WebP), the files are replaced 1:1 after the rebuild.

So maybe this should be a hint somewhere

@Chris D

 

[rdn]

I would not try to use it with PHP 8.4 as it produces Zend memory errors that point to some problems

Please add more information about this?
I've been using PHP 8.4 since it was released for several XF 2.3 servers I managed without any issues.

 

[Mbx]

Please add more information about this?
I've been using PHP 8.4 since it was released for several XF 2.3 servers I managed without any issues.

Sure. I try.

The encountered problems are connected to the interaction between OpCache and the core-compiled.js script, only noticeable when users submit new posts on the Xenforo platform. This situation results in confusing error messages for the end-users.

Also, an unusual increase in interned string memory usage was observed. After the update, memory consumption rose sharply from ~ 8M to ~ 50M.


.


When available memory dropped to zero, occasionally triggering massive but randomly "Bad Gateway" errors.

Following this, a similar issue surfaced less frequently, but with an added complication: a "zend_mm_heap corrupted" error. Rather than conducting extensive debugging, this problem was presumed to stem from an incompatibility. This assumption was later confirmed.

An upgrade to version 8.3 was implemented, which resolved these issues without requiring any further changes to the configuration.

These errors seemed more / only prevalent on mobile devices. Board size might be a factor as well. But not 100% sure, just points to check.

So its a little bit tricky. I would not recommend 8.4 as for now.

 

[rdn]

These errors seemed more prevalent on mobile devices.

It seems odd to me that a specific PHP version would be particularly problematic for mobile devices, given that the same backend code is also used for desktop users.

Do you have some server error logs related to these errors?
XenForo admin log, nginx, or PHP error logs.

I've been running PHP 8.4 without issues since 8.4.1 was released.

 

[Mr Lucky]

saving the WebP files separately (which would been better, like for checking support and then choosing between JPEG and WebP),

I agree that would be a good idea, but I’m fine with 55% webp anyway.

 

[rdn]

From your previous image attached (now deleted), that a JavaScript URI file seems to contain unnecessary or extra parameters seems to be caused by the webserver rewrites, incompatible addons, or some outdated templates.

 

[Mbx]

It seems odd to me that a specific PHP version would be particularly problematic for mobile devices, given that the same backend code is also used for desktop users.

Do you have some server error logs related to these errors?
XenForo admin log, nginx, or PHP error logs.

I've been running PHP 8.4 without issues since 8.4.1 was released.

No coding errors. I can only say its related to OpCache and the JS process for sure. Maybe a memory leak.

Another thought was -- CPU optimizations might be something to check; used O3 with latest instructions (maybe too aggressive).

So I am not sure if this is not related to Php 8.4 itself and/or maybe only happen for special setups in regards to Xenforo. Board size might be a factor ^ as commented before. For mobile; its notable there, I suppose, because the JS functions might work differently for desktop, not leading to the board's error message, while the error still happens. The error does not prevent posting, it just requires a reload, causing many users to double post + bad user experience.

If its a memory leak, which I assume, more likely >= big board. And ours is very large.

 

[Mbx]

I agree that would be a good idea, but I’m fine with 55% webp anyway.

Well, I did not care so much as well. Speed vs Quality -- but the users. So I think, 90 % serves both interests.