Running server with http_unix_daemon does not respond very well to signals

[rla]

I compiled SWI-Prolog 8.5.16 using the standard instructions (Debian 11).

Test code:

:- use_module(library(http/http_json)).
:- use_module(library(http/http_unix_daemon)).
:- use_module(library(http/thread_httpd)).

http_unix_daemon:http_server_hook(Options):-
    http_server(reply, Options).

reply(_):-
  reply_json(_{ status: hello }).

Command line:

swipl test.pl --no-fork --port=15043 --workers=8 --interactive=false

Sending SIGINT does nothing except in 5 or so minutes where it throws:

ERROR: 2022-09-15 22:20:00.000 Stack limit (1.0Gb) exceeded
ERROR: 2022-09-15 22:20:00.000   Stack sizes: local: 11Kb, global: 3Kb, trail: 0Kb
ERROR: 2022-09-15 22:20:00.000   Stack depth: 30, last-call: 33%, Choice points: 13
ERROR: 2022-09-15 22:20:00.000   In:
ERROR: 2022-09-15 22:20:00.000     [30] '$messages':longest_prefix([], _1028)
ERROR: 2022-09-15 22:20:00.000     [29] '$messages':longest_prefix([length:1], [length:1|_1060])
ERROR: 2022-09-15 22:20:00.000     [28] '$messages':longest_id(en_US, _1080)
ERROR: 2022-09-15 22:20:00.000     [26] '$messages':translate_message(<compound debug/2>, _1100, [])
ERROR: 2022-09-15 22:20:00.000     [25] '$dcg':call_dcg(<compound (:)/2>, _1128, [])
ERROR: 2022-09-15 22:20:00.000 
ERROR: 2022-09-15 22:20:00.000 Use the --stack_limit=size[KMG] command line option or
ERROR: 2022-09-15 22:20:00.000 ?- set_prolog_flag(stack_limit, 2_147_483_648). to double the limit.

Removing thread_idle wrapper from wait/1 in https://github.com/SWI-Prolog/packages-http/blob/4a1cc0aa520dfe6244831fc60e1756afaf4cd6ea/http_unix_daemon.pl#L837 fixes it. Signal seems to be sent but the thread seems to have trouble receiving it when thread_idle is used.

[JanWielemaker]

Thanks. The thread_idle call was unrelated :smile: Fixed with SWI-Prolog/swipl-devel@8c9d3757bf900403e8b6428e21bf4f980061d70a

[rla]

Thanks for the quick fix!