[asterisk-dev] AST_MODULE_LOAD_FAILURE vs AST_MODULE_LOAD_DECLINE

[Corey Farrell]

Probably not the popular opinion, but I don't think any modules are 
"generally" important enough to abort Asterisk start-up.  For any module 
that is important enough for a system the admin can use 'require' or 
'preload-require' settings in modules.conf.  The exception to this would 
be when a module creates an unstable state by being half-loaded 
(example: global symbols exported that will crash if called).

On the 'core' side of things, anytime main/loader.c aborts start-up it 
should display an error message saying which module returned 
AST_MODULE_LOAD_FAILURE and make sure the logger gets flushed.  Even if 
every module gives a reason it is failing the module loader should still 
print a message (two errors on abort are better than possibly missing one).


On 04/05/2017 04:43 PM, George Joseph wrote:
> Over the years I've been frustrated at times where Asterisk fails to 
> start for absolutely no (apparent) reason.  No error message, no trap, 
> nothing.  It just ends.  Every case of this I've tracked down has been 
> the result of a module returning AST_MODULE_LOAD_FAILURE when it 
> encounters a problem but not bothering to print an error first.   If 
> you don't know already, AST_MODULE_LOAD_FAILURE tells asterisk to 
> *stop loading and exit*.  What the module should have done was 
> actually print an error and return AST_MODULE_LOAD_DECLINE which just 
> tells asterisk that the module couldn't load.
>
> So now I'm doing an audit of module return codes to see which call 
> FAILURE vs DECLINE and found the modules below all call FAILURE.
>
> The big question is...  Exactly which modules are so critical to 
> asterisk that asterisk should exit if the module couldn't load?  
> Surely not all 103 modules listed below.  If we can agree on the ones 
> that really /should/ call FAILURE, I'll update those with error 
> messages, then change the rest to DECLINE and add error messages as 
> needed.
>
> Reply with the list intact.  Don't add, remove, split or re-order the 
> lines.  Just stick an "F " before the modules you think should remain 
> failure.
>
> Modules returning FAILURE:
>
> ./addons/app_mysql.c
> ./addons/cdr_mysql.c
> ./addons/chan_mobile.c
> ./apps/app_adsiprog.c
> ./apps/app_agent_pool.c
> ./apps/app_alarmreceiver.c
> ./apps/app_amd.c
> ./apps/app_authenticate.c
> ./apps/app_cdr.c
> ./apps/app_confbridge.c
> ./apps/app_festival.c
> ./apps/app_followme.c
> ./apps/app_forkcdr.c
> ./apps/app_meetme.c
> ./apps/app_queue.c
> ./apps/app_skel.c
> ./apps/app_voicemail.c
> ./cdr/cdr_custom.c
> ./cel/cel_custom.c
> ./cel/cel_odbc.c
> ./channels/chan_alsa.c
> ./channels/chan_console.c
> ./channels/chan_dahdi.c
> ./channels/chan_iax2.c
> ./channels/chan_mgcp.c
> ./channels/chan_misdn.c
> ./channels/chan_motif.c
> ./channels/chan_nbs.c
> ./channels/chan_oss.c
> ./channels/chan_phone.c
> ./channels/chan_pjsip.c
> ./channels/chan_sip.c
> ./channels/chan_skinny.c
> ./channels/chan_unistim.c
> ./channels/chan_vpb.cc
> ./codecs/codec_a_mu.c
> ./codecs/codec_adpcm.c
> ./codecs/codec_alaw.c
> ./codecs/codec_g722.c
> ./codecs/codec_g726.c
> ./codecs/codec_gsm.c
> ./codecs/codec_ilbc.c
> ./codecs/codec_lpc10.c
> ./codecs/codec_resample.c
> ./codecs/codec_ulaw.c
> ./formats/format_g723.c
> ./formats/format_g726.c
> ./formats/format_g729.c
> ./formats/format_gsm.c
> ./formats/format_h263.c
> ./formats/format_h264.c
> ./formats/format_ilbc.c
> ./formats/format_jpeg.c
> ./formats/format_ogg_vorbis.c
> ./formats/format_pcm.c
> ./formats/format_sln.c
> ./formats/format_vox.c
> ./formats/format_wav.c
> ./formats/format_wav_gsm.c
> ./funcs/func_cdr.c
> ./main/loader.c
> ./pbx/pbx_dundi.c
> ./pbx/pbx_loopback.c
> ./pbx/pbx_realtime.c
> ./pbx/pbx_spool.c
> ./res/res_pjsip/config_transport.c
> ./res/res_ari.c
> ./res/res_ari_events.c
> ./res/res_ari_model.c
> ./res/res_calendar.c
> ./res/res_chan_stats.c
> ./res/res_clialiases.c
> ./res/res_config_ldap.c
> ./res/res_config_sqlite.c
> ./res/res_config_sqlite3.c
> ./res/res_curl.c
> ./res/res_endpoint_stats.c
> ./res/res_fax.c
> ./res/res_hep_rtcp.c
> ./res/res_http_websocket.c
> ./res/res_musiconhold.c
> ./res/res_odbc.c
> ./res/res_phoneprov.c
> ./res/res_pjsip_nat.c
> ./res/res_pjsip_one_touch_record_info.c
> ./res/res_pjsip_outbound_publish.c
> ./res/res_pjsip_outbound_registration.c
> ./res/res_pjsip_sdp_rtp.c
> ./res/res_pjsip_send_to_voicemail.c
> ./res/res_pjsip_t38.c
> ./res/res_smdi.c
> ./res/res_snmp.c
> ./res/res_stasis.c
> ./res/res_stasis_device_state.c
> ./res/res_stasis_playback.c
> ./res/res_stasis_recording.c
> ./res/res_stasis_test.c
> ./res/res_statsd.c
> ./res/res_timing_kqueue.c
> ./res/res_xmpp.c
> ./res/res_pjsip_pubsub.c
> ./tests/test_bucket.c
> ./tests/test_channel_feature_hooks.c
>
> -- 
> George Joseph
> Digium, Inc. | Software Developer
> 445 Jan Davis Drive NW - Huntsville, AL 35806 - US
> Check us out at: www.digium.com <http://www.digium.com/> & 
> www.asterisk.org <http://www.asterisk.org/>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20170405/2e6ebd46/attachment-0001.html>