Em 23/11/2016 11:54, Adriano dos Santos Fernandes escreveu: > On 23/11/2016 11:50, Alexandre Benson Smith wrote: >> Em 23/11/2016 11:37, Adriano dos Santos Fernandes escreveu: >>> And in my opinion problem of metadata changes and >>> unreliable/unpredictable behavior with concurrent sessions precedes this >>> one, otherwise deeper recompilation will be a big problem. >> Perhaps I missed your point. > I don't.
It's good since at least one of us still on the track :) >> The case I provided has no concurrent sessions and persists even after a >> reconnect cycle. >> > But that (DDL changes with concurrent sessions) is unreliable. It's unreliable with no concurrent sessions either. Since the view that uses a sub-select or a column computed by the sub-select of that field trows an error in production without any warning during the data type change. > > And if you change a single view that recompiles 100 views, you multiple > the unreliability by 100 in the production environment. So, the only viable option is to prohibit any change that has dependencies as it was on FB 2.1 and earlier ? > > Adriano ------------------------------------------------------------------------------ Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
