https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=43911 --- Comment #4 from Martin Renvoize (ashimema) <martin.renvoize@openfifth.co.uk> --- Thanks David, very good catch, and thank you for noting what the tests leave behind. The failure was in the tests rather than the fix: balanced_reconciliations (and the other cashup tests) set up CashupPaymentTypes themselves, with CASH and CHEQUE, but not the library limitations of the PAYMENT_TYPE authorised values. On a system where CHEQUE is limited to one library (as after your manual test plan) the test register, which is created in a random library, correctly no longer lists CHEQUE when it has no takings of it, so the test saw one entry instead of two. On a database where CHEQUE has no limitation they pass, which is why they passed for you before the reset. I reproduced it with CHEQUE limited to one library and the new attachment makes the tests remove the limitations inside their transactions. They now pass with and without such a limitation. The library and cash register left behind came from the cashup_payment_types test (from bug 33501) which created a register outside of a transaction; that is fixed in the same patch and I checked the counts of libraries and registers no longer change after a run. Setting back to Needs Signoff. -- You are receiving this mail because: You are watching all bug changes.