albertlast

the db length depend if the database run in mb4 mode or not, atm mysql false and pg true. but in near future this behavior is mysql variable: #6406 #6409

Sonne kind of test data would be nice, To had better understanding of the this.

well this say less than you believe, since this place is only the last place who called for more memory, so other places could consume the main memory. But since...

this stuff get rebuild when a change happen right?

``GROUP_CONCAT(xpv.id_group ORDER BY xpv.id_group)`` = ``string_agg(xpv.id_group, ',' ORDER BY xpv.id_group)`` this query could be very slow, so question: would be better instead of doing a check, just run rebuild on...

in general when it find and error than it should rebuild complet, this is the default behavior of smf on changes and this should be also the same for repair,...

Well i try this already by my own, to get an idea how this could work. End of the day like you mention is the access of odata/rest, the point...

i would like to stick the scope of this pr to news.

Could it be, that you implement the pr #6893 and so deleted the approved index on the message table? this index by his own shoujld improved your performance.

got no idea how to fixit for mysql