[openbsd] server crash on pickDeliver/pickDeliver/innerQuery.sql and vrp_basic/no_crash_test-vrpOneDepot.sql

[landryb]

Describe the bug
running tests on OpenBSD, two tests crash the remote server (figured out the two after removing them selectively, all tests pass, per openbsd/ports@d4bd6a5) - cf #1356 (comment)

that's with head of master and head of release-3.0 branch.

../../pgtap/pickDeliver/pickDeliver-VS-pickDeliverEuclidean.sql ........................................ ok            
psql:../../pgtap/pickDeliver/pickDeliver/innerQuery.sql:285: server closed the connection unexpectedly                                                                                                                                         
        This probably means the server terminated abnormally                                                           
        before or while processing the request.                                                                        
psql:../../pgtap/pickDeliver/pickDeliver/innerQuery.sql:285: fatal: connection to server was lost                                                                                                                                              
../../pgtap/pickDeliver/pickDeliver/innerQuery.sql .....................................................               
Dubious, test returned 2 (wstat 512, 0x200)                                                                            
Failed 45/77 subtests                                                                                                                                                                                                                          
psql: error: could not connect to server: FATAL:  the database system is in recovery mode                              
../../pgtap/pickDeliver/pickDeliver/types_check.sql ....................................................               
Dubious, test returned 2 (wstat 512, 0x200)                                                                                                                                                                                                    
No subtests run  

[landryb]

same failure on develop branch.

[cvvergara]

@landryb
Maybe this issue is related with #1760
There is code that pgr_TSP & pgr_PickDeliver share
I wonder if you can try 2 things:

1. Test the code posted on the issue TSP server crash on ubuntu 20.04 #1760 on openbsd and see if you are getting the same error
2. Test if with the fix I am working for TSP server crash on ubuntu 20.04 #1760 also fixes this issue, you can find it here: https://github.com/cvvergara/pgrouting/tree/issue-1760-3.1

[landryb]

tried with cvvergara/pgrouting@a62cd7801 backported on top of 3.1.1 but alas the test still fails:

../../pgtap/pickDeliver/pickDeliver-VS-pickDeliverEuclidean.sql ........................................ ok                                                                                                                                    
psql:../../pgtap/pickDeliver/pickDeliver/innerQuery.sql:322: server closed the connection unexpectedly                                                                                                                                         
        This probably means the server terminated abnormally                                                                                                                                                                                   
        before or while processing the request.                                                                                                                                                                                                
psql:../../pgtap/pickDeliver/pickDeliver/innerQuery.sql:322: fatal: connection to server was lost 

fwiw that's with clang 10.0.1

[landryb]

oh wait, i think i tested that wrong. retesting.

[landryb]

well, smth feels wrong because now it fails on the previous test (unless tests are run several at a time in parallel ?):

../../pgtap/mincut/stoerWagner-types-check.sql ......................................................... ok                                                                                                                                    
../../pgtap/pickDeliver/order_id_is_neg1.sql ........................................................... ok   
psql:../../pgtap/pickDeliver/pickDeliver-VS-pickDeliverEuclidean.sql:33: server closed the connection unexpectedly                                                                                                                             

[cvvergara]

@landryb
Can you try this one please
https://github.com/cvvergara/pgrouting/tree/trying-to-fix-pick-deliver-crash

[landryb]

build from a tarball taken from this branch (eg https://github.com/cvvergara/pgrouting/archive/bf23f07e3f9bdc97ba184e61ae012b3c63d30948.tar.gz but thats based off 3.0 branch, right ?) but sadly that fails the same:

../../pgtap/pickDeliver/pickDeliver-VS-pickDeliverEuclidean.sql ........................................ ok            
psql:../../pgtap/pickDeliver/pickDeliver/innerQuery.sql:273: server closed the connection unexpectedly                                                                                                                                         
        This probably means the server terminated abnormally                                                                                                                                                                                   
        before or while processing the request.
psql:../../pgtap/pickDeliver/pickDeliver/innerQuery.sql:273: fatal: connection to server was lost

which particular commit did you have in mind from that branch that should fix this crash ?

[cvvergara]

Well, on CENTOS there was a problem with boost version so I added external boost for geometries I was thinking that that might have been the problem, v3.0.0-alpha does not have those external geometries. I was thinking that that was the source of the problem

git diff --name-only v3.0.0-alpha v3.0.0 include
include/boost/bgeometry/algorithms/append.hpp
include/boost/bgeometry/algorithms/area.hpp
include/boost/bgeometry/algorithms/assign.hpp
include/boost/bgeometry/algorithms/centroid.hpp
include/boost/bgeometry/algorithms/clear.hpp
include/boost/bgeometry/algorithms/convert.hpp
include/boost/bgeometry/algorithms/correct.hpp
include/boost/bgeometry/algorithms/covered_by.hpp
...

[cvvergara]

Don't know exactly how yet, but basically working on removing warnings from clang as a compiler, I could finally duplicate a server crash with pgr_pickDeliver on ubuntu
https://github.com/cvvergara/pgrouting/runs/1493551654?check_suite_focus=true#step:9:405
Note:
Before trying to attempt to fix the issue I need to finish removing clang's warnings. hopefully nothing else will come up.

[cvvergara]

@landryb
Can you try this one please
https://github.com/cvvergara/pgrouting/tree/dec-11-fix-server-crash-on-pickDeliver-for-3.1

Please let me know if it works, so that I start cleaning all the code mess I created to find & fix the server crash.

[landryb]

victory ! 🎉

All tests successful.

Test Summary Report
-------------------
../../pgtap/alpha_shape/s_shape-test.sql                                                             (Wstat: 0 Tests: 167 Failed: 0)
  TODO passed:   95
../../pgtap/trsp/trsp/innerQuery.sql                                                                 (Wstat: 0 Tests: 162 Failed: 0)
  TODO passed:   4-6, 10-12, 16-18, 34-36, 40-42, 46-48
                58-60, 64-66, 70-72, 88-90, 94-96, 100-102
                112-114, 118-120, 124-126, 142-144, 148-150
                154-156
../../pgtap/trsp/trspViaEdges/innerQuery.sql                                                         (Wstat: 0 Tests: 162 Failed: 0)
  TODO passed:   4-6, 10-12, 16-18, 34-36, 40-42, 46-48
                58-60, 64-66, 70-72, 88-90, 94-96, 100-102
                112-114, 118-120, 124-126, 142-144, 148-150
                154-156
../../pgtap/trsp/trspViaVertices/innerQuery.sql                                                      (Wstat: 0 Tests: 162 Failed: 0)
  TODO passed:   4-6, 10-12, 16-18, 34-36, 40-42, 46-48
                58-60, 64-66, 70-72, 88-90, 94-96, 100-102
                112-114, 118-120, 124-126, 142-144, 148-150
                154-156
Files=302, Tests=32094, 595 wallclock secs ( 5.31 usr  5.31 sys +  1.89 cusr  8.75 csys = 21.26 CPU)
Result: PASS

[cvvergara]

Ok, perfect will proceed to clean up