Update xorm

[KN4CK3R]

Noticed in #25613 that the new xorm version does work with Gitea. Created this PR to keep the other one clean.

The method with the error is models/unittest.GetCount changed in #24963. This method generates a query like

SELECT  count(*) FROM "gtestschema"."user" WHERE (type = 0) ORDER BY id

which is invalid for pgsql and mssql:
ERROR: column "user.id" must appear in the GROUP BY clause or be used in an aggregate function
The error message is a bit strange but for COUNT the order should be irrelevant.
The old xorm version generates the query without ORDER BY

SELECT count(*) FROM "gtestschema"."user" WHERE (type = 0)

@lunny fyi

[lunny]

Yes. It's a break change of xorm new version. Previously, count with order by will ignore the order by statement automatically. But the new xorm version will not ignore it automatically. Users should handle themselves.

For Gitea, when do count, order by should not be appended.

[lunny]

Should I revert back the change to ignore order by when it's counting?

[KN4CK3R]

No, I think it's fine if we need to write correct queries. But how does it work with FindAndCount. The find part should use the order and the count part should not. If xorm does ignore it in this case it's inconsistent handling which may be bad too.

[lunny]

No, I think it's fine if we need to write correct queries. But how does it work with FindAndCount. The find part should use the order and the count part should not. If xorm does ignore it in this case it's inconsistent handling which may be bad too.

Yes, I have thought of that. And that's why I think maybe it's need to be reverted to original logic.

[KN4CK3R]

Yes, then the old logic may be better.

[lunny]

Yes, then the old logic may be better.

Created a new PR to try to fix it https://gitea.com/xorm/xorm/pulls/2307

[lunny]

https://gitea.com/xorm/xorm/pulls/2307 merged.

[KN4CK3R]

Seems worse now?

[lunny]

Seems worse now?

I found this is a bug of Gitea's test because Gitea use an empty string as OrderBy's parameter.

[lunny]

This could be replaced by #26128