Pavel Luzanov <p.luza...@postgrespro.ru> writes: >> I wonder if we shouldn't put these new tests in largeobject.sql instead. >> (That would mean there are two expected-files to change, which is a bit >> tedious, but it's not very hard.)
> I made the same changes to two files in the 'expected' directory: > largeobject.out and largeobject_1.out. > Although I must say that I still can't understand why more than one file > with expected result is used for some tests. Because the output sometimes varies across platforms. IIRC, the case where largeobject_1.out is needed is for Windows, where the file that gets inserted into one of the LOs might contain CR/LF not just LF newlines, so the LO contents look different. > Also, I decided to delete following line in the listLargeObjects > function because all the other commands in describe.c do not contain it: > myopt.topt.tuples_only = false; Agreed, I'd done that already in my version of the patch. > Second version (after splitting) is attached. Pushed with some minor editorialization. regards, tom lane