| moving to wallet-core, i will add more logging to know what's happening | |
| 86194c614 * adds more log in wallet core when database open fails * uses the standard cause field https://developer.mozilla.org/es/docs/Web/JavaScript/Reference/Global_Objects/Error/cause * enhance Error.toString to print cause information in logs * save walletInit failure to printout in UI we need to also implement a logging feature to export logs | |
| waiting for feedback | |
| Emmanuel, can you please try again and provide feedback? | |
| From the feedback, the error is "hint": "wallet core not available, last error: Error: database opening error\n Caused by: InvalidStateError: A mutation operation was attempted on a database that did not allow mutations." We have seen this already in https://bugs.gnunet.org/view.php?id=7777 I was not able to produce this in firefox 123 (current latest) nor in 115 (my current) and 104 (the oldest i have) BUT in 122 there is a new option "always use private browsing mode" which funny enough it is not present in the latest version. IF i'm using 122 and have this options turned on then I can reproduce the issue. This also explain why a new profile fix the problem (the options is off by default). Seems like a firefox problem, upgrading to 123 may help. Selection_301.png (27,252 bytes) Selection_302.png (15,429 bytes) | |
| related https://bugzilla.mozilla.org/show_bug.cgi?id=1427986 | |
| Can we please show a proper error message? Like I think wallet-core should *start* (not fail hard), and then the first API call to wallet-core (that usually returns the version data!) needs to return a nice "private browsing mode firefox buggy database access denied" error message. The GUI should then just render the error message (whatever it is). So wallet-core needs to 1) be modified to not exit, even on fatal errors 2) be extended with a general 'fatal error' protocol message to be returned to the GUIs Right? | |
| Even right now, wallet-core does not "exit" on fatal errors. When any wallet-core request fails, the error is returned to the caller via an error notification. If any task fails, the result is stored in the DB (at least an attempt is made ...) and we re-try. However in this situation, it looks like wallet-core isn't even responding to the "init" call. Sebastian, can you check why this is happening? | |
| "the body cannot live without the mind" said in a famous movie, and here "wallet cannot live without a database" initWallet is already doing database write op assuming database is ready and healthy, it needs to check and return a taler error code if there is not database, or the database is read only. seems like the "permanent private browsing" mode in firefox will fall under "can't access to database" type | |
| wallet core added more info when db is not ready commit 8dcdcb9a7df3838e2b3ba4ef2d50e22126709ec6 (HEAD -> master, origin/master, origin/HEAD) Author: Florian Dold <florian@dold.me> Date: Tue Feb 20 19:57:11 2024 +0100 wallet-core: better error handling for DB issues in init request | |
| not enough, wallet breaks before trying to write... i'm still getting 7011 (wallet unavailable) | |
| looks like the problems is in ensureWalletDbOpen, i will add an catch there wallet-private-mode-error.txt (1,135 bytes) | |
| now this can be reproduce with FIREFOX_PATH=/path/to/firefox/122.0.1 pnpm test:firefox-private | |
| now error is propagated and UI catch it. if the problem is DB_UNAVAILABLE the shown message is "Could not open the wallet database" but if this happen in firefox and private mode the next error will be shown image.png (25,706 bytes) | |
| 6c124075b | |
| Are you showing just an incomprehensible error, or did you 1) either figure out how to make it work despite FF in private mode, or 2) are giving a *good* actionable error message that users can work with and fix the issue? Because merely propagating the error isn't quite fully addressing the specific issue... | |
- Anonymous