>> The newly added PSQL_WATCH_PAGER feature which broke the patch does not >> seem to be tested anywhere, this is tiring:-( > > Do you have any idea how this can be tested?
The TAP patch sent by Peter on this thread is a very good start.
> It requires some pager that doesn't use blocking reading, and you need > to do remote control of this pager. So it requires a really especially > written pager just for this purpose. It is solvable, but I am not sure > if it is adequate to this patch.
Not really: The point would not be to test the pager itself (that's for the people who develop the pager, not for psql), but just to test that the pager is actually started or not started by psql depending on conditions (eg pset pager…) and that it does *something* when started. See for instance the simplistic pager.pl script attached, the output of which could be tested. Note that PSQL_PAGER is not tested at all either. Basically "psql" is not tested, which is a pain when developing a non trivial patch.
Minimally for PSQL_WATCH_PAGER, the pager should exit after some time, but before it has to repeat data reading. Elsewhere the psql will hang.
can be solution to use special mode for psql, when psql will do write to logfile and redirect to file instead using any (simplified) pager? Theoretically, there is nothing special on usage of pager, and just you can test redirecting to file. That is not tested too. In this mode, you can send sigint to psql - and it can be emulation of sigint to pager in PSQL_WATCH_PAGER mode,