search

SWI-Prolog / packages-pengines

Pengines: Prolog engines
11 stars 13 forks source link

The id option of pengines_create doesn't work #21

Closed torbjornlager closed 8 years ago

torbjornlager commented 8 years ago

This used to work, but no longer does:

?- use_module(library(pengines). true. ?- pengine_create([id(Pid)]). ERROR: thread_create/3: Uninstantiated argument expected, found [id(_G1151)](2-nd argument)

JanWielemaker commented 8 years ago

Which version are you running? Works just fine here (7,.3.15, but this hasn't changed in a while)

1 ?- use_module(library(pengines)).
true.

2 ?- pengine_create([id(Pid)]).
Pid = '96091231-e33c-4dec-b5d1-66099ef11887'.
torbjornlager commented 8 years ago

Hmm, seems to work for me too now. But I think I know what fooled me into thinking it didn't work at all. It appears that if I create three pengines in a row, the behaviour is less than helpful if I try to create a forth one, when the slave-limit kicks in. See below. So not really a bug, but a behaviour that could perhaps be made more helpful.

?- pengine_create([id(Pid)]). Pid = '4d079347-1d40-480e-b10e-826561861c39'.

?- pengine_create([id(Pid)]). Pid = '8be95daa-95a6-4e9c-99f1-a55ab1d4d411'.

?- pengine_create([id(Pid)]). Pid = 'fbb7aa9e-56a2-4c89-af4f-f797ee1c655a'.

?- pengine_create([id(Pid)]). true. <== NOT HELPFUL

?- thread_get_message(M). M = pengine_event('4d079347-1d40-480e-b10e-826561861c39', create('4d079347-1d40-480e-b10e-826561861c39', [slave_limit(3)])).

?- thread_get_message(M). M = pengine_event('8be95daa-95a6-4e9c-99f1-a55ab1d4d411', create('8be95daa-95a6-4e9c-99f1-a55ab1d4d411', [slave_limit(3)])).

?- thread_get_message(M). M = pengine_event('fbb7aa9e-56a2-4c89-af4f-f797ee1c655a', create('fbb7aa9e-56a2-4c89-af4f-f797ee1c655a', [slave_limit(3)])).

?- thread_get_message(M). M = pengine_event('d6105145-3b52-4b4d-b182-1d2977cbec98', error('d6105145-3b52-4b4d-b182-1d2977cbec98', error(resource_error(max_pengines), _G1476))).

JanWielemaker commented 8 years ago

Changed to make pengine_create/1 return the create error as an exception. I think that makes more sense. Strictly not really a bug though. I can even imagine we can think of scenarios where the old behaviour is desired. Well, we can add an option for that case.