| 16 Feb 2019 |
 Linda | mmhm | 00:09:19 |
| Linda | With matrix-archive-keyring, I've made sure debconf(1) remembers the answers to questions and doesn't prompt unnecessarily, even if the caches are gone.
+debian0.10 version atm.
| 00:10:06 |
| Linda | Probably isn't in the +debian0.6 version I gave you | 00:10:26 |
| Linda | so it may forget from time to time | 00:10:33 |
 Maximus | I'm going to do a purge and full re-install anyway | 00:12:07 |
| Maximus | clearly this is not right and I want a clean setup | 00:12:21 |
| Maximus | can't have stats reporting on my managed servers... | 00:12:33 |
| Maximus | or even risk having it | 00:12:49 |
| Linda | Well, as it currently is (you're right with reporting the issue), it will keep reactivating. | 00:15:17 |
| Linda | I can give you an updated source/deb | 00:15:28 |
| * Maximus sighs | 00:16:29 |
| Maximus | Linda: thank you, but it's ok. I'll just lock the file so it can't be updated anymore | 00:17:04 |
 [uhoreg] | In reply to @lindalap:matrix.org
- If there is no answer to the question, it asks. You choose no.
- The answer is "false" this time.
report_stats is not enabled, debconf(1) remembers this question in its cache.
- Next time you upgrade or reconfigure
matrix-synapse-py3, debconf(1) checks its cache for answer. If it is there, $RET will exit with a status code 10 or higher. Or effectively "true".
You can see where this goes. I don't get this. $RET gets filled with the value of the report-stats option, so the config file gets written with whatever report-stats in set to. It also doesn't explain why debconf has the option set to true on Max's system. | 00:17:14 |
| Linda |
dpkg-buildpackage: warning: build dependencies/conflicts unsatisfied; aborting
Well, guess I wouldn't do that anyway | 00:17:55 |
| Maximus | uhoreg: because when I installed it the very first time, it was unattended. It defaulted to true back then. and I guess I never updated the cache | 00:18:00 |
| Maximus | but regardless: the report_stats.yaml file exists. it must not be overwritten | 00:18:22 |
| Maximus | doesn't matter what the cache is | 00:18:32 |
| Linda | uhoreg: Because of db_input high matrix-synapse/report-stats || true? Actually, true being exit code 1. It satisfies [ if "$RET" ]; then. | 00:18:42 |
| Linda | || true happens if there's a known answer.
| 00:18:56 |
| Linda | Actually, uhoreg might be right | 00:19:39 |
| Maximus | in any case, the change happened on a system where the cache was false, and the config file itself was set to false | 00:21:54 |
| Maximus | just discovered it on another server | 00:22:03 |
| Maximus | so it can go to true either way | 00:22:12 |
| [uhoreg] | The || true is just there so that the script doesn't die if the db_input command errors for whatever reason. It doesn't affect the behaviour of db_input. | 00:23:27 |
| Linda | It's a false assumption $RET will be "true" or "false for a boolean question. | 00:23:28 |
| Linda | My brain hurts, I don't want to think more about it | 00:24:10 |
| Linda | Even debconf-devel(7) man page implies to check "true" or "false". | 00:24:59 |
| Linda | asfsaag | 00:25:08 |
| Linda | stop, Linda | 00:25:09 |
| Maximus | yes Linda, stop and enjoy your weekend! | 00:26:11 |
