Hey Andres, > Awesome! +1. Attached 2nd version of patch rebased on master. > (v2-0001-Resolve-PL-handler-names-for-JITed-code-instead-o.patch) > > > > > Did you check whether there's any cases this fails in the tree with your > > patch applied? The way I usually do that is by running the regression > > tests like > > PGOPTIONS='-cjit_above_cost=0' make -s -Otarget check-world > > > > (which will take a bit longer if use an optimized LLVM build, and a > > *lot* longer if you use a debug llvm build) > > > > Great suggestion! I used: > PGOPTIONS='-c jit_above_cost=0' gmake installcheck-world > It all passed except a couple of logical decoding tests that never pass > on my machine for any tree (t/006_logical_decoding.pl and > t/010_logical_decoding_timelines.pl) and point (which seems to be failing > even > on master as of: d80be6f2f) I have attached the regression.diffs which > captures > the point failure.
I have attached the 3rd version of the patch rebased on master. I made one slight modification to the previous patch. PL handlers, such as that of plsh, can be in an external library. So I account for that in modname (earlier naively I set it to NULL). There are also some minor changes to the comments and I have rehashed the commit message. Apart from running the regress tests as you suggested above, I installed plsh and forced JIT on the following: CREATE FUNCTION query_plsh (x int) RETURNS text LANGUAGE plsh AS $$ #!/bin/sh psql -At -c "select 1" $$; SELECT query_plsh(5); and I also ran plsh's make installcheck with jit_above_cost = 0. Everything looks good. I think this is ready for another round of review. Thanks!! Soumyadeep
v3-0001-Resolve-PL-handler-names-for-JITed-code-instead-o.patch
Description: Binary data
