I'm using LW 4.2.6 on Linux, and have noticed that sql calls block my process - no scheduling takes place before the sql calls returns. Not very surprising when doing FFI - but is there a simple way to avoid it? (I'm thinking of writing an extra server layer between my server and my database, but for all I know there could be simpler ways...) -- (espen)
>>>>> Espen Vestre writes: Espen> I'm using LW 4.2.6 on Linux, and have noticed that sql calls block Espen> my process - no scheduling takes place before the sql calls Espen> returns. Not very surprising when doing FFI - but is there a Espen> simple way to avoid it? (I'm thinking of writing an extra server Espen> layer between my server and my database, but for all I know there Espen> could be simpler ways...) I've been fighting this problem for nearly two years. Does anybody know if one could write a specialized version of unixODBC which advertises which socket (fd) is being used to communicate to the back end, so that this socket included in the lisp scheduler's list of noticed fds? Would that work? Would one also have to re-write the unixODBC driver to not do any blocking calls in this case? (i.e. check for EWOULDBLOCK)? -- Alain Picard Memetrics
Alain Picard <Alain.Picard@memetrics.com> writes: > I've been fighting this problem for nearly two years. > > Does anybody know if one could write a specialized version > of unixODBC which advertises which socket (fd) is being used fyi: I'm using the oracle libraries, not odbc -- (espen)