-
-
Notifications
You must be signed in to change notification settings - Fork 360
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Feature Request / Bug] Don't process.exit(1) in the API; Throw an error instead #516
Comments
No need to fix. Just tell db-migrate to not register the events on uncatched{Rejection,Exception}. The options property you need to set for this is, Line 42 in 4de822b
|
Ah and I just saw the createMIgrationsDir example, that should be probably just removed though. |
Thanks so much! ❤️ I added a PR to the docs to add this option. (If that was the wrong place, let me know). Thanks for the help! |
My pleasure, and thank you for contributing on the docs <3 I will also push out a change to replace the aforementioned exit. |
<3 |
Hats off to @Flaque's diplomacy skills. I, on the other hand, am pretty speechless. Because I just spent a number of minutes figuring as to why the following code (async () => {
const dbmigrate = require('db-migrate')
process.env.DATABASE_URL = 'postgres://localhost:5432'
const dbm = dbmigrate.getInstance(true)
await dbm.dropDatabase('pr-apps-test')
await dbm.createDatabase('pr-apps-test')
await dbm.up()
})().catch(e => {
console.error(e)
}) And what do I find in the end? |
:'( Sorry you ran into that @artemave. Glad you found the issue though! |
@artemave db-migrate is not a library, but lately also a library ;) So for programmatic modes, it could happen that some stuff should not be where it is. I am not sure though, what you wanted to reach with your comment here. Which is not a bug report, nor a feature suggestion, nor you seem to want to help out or something. It feels more like you just wanted to brag around here, I am sorry if you did not wanted to, but this clearly reads like it. I want to friendliest point yout to the COC. So you did ran into the issue of public registers right? Can't say since you did not provided a proper problem description, though. Just correct me should I am be wrong. However, when @Flaque opened this issue, I'm not quite sure if I respected the programmatic mode at that time, I agree however, that for that mode we should default to not register the error handlers. |
A library must not:
This library does both and that is about as most accurate description of the problem as I can come up with. |
@artemave DB-Migrate was once just a cli tool and now started to provide abilities to be used as a module. You're doing nothing more than wasting time, if you want to actually help the open source community, be constructive and helpful. Yes bugs exists, problems exist and all is fine. But I see I was right, you do want only brag around here, and with that I kindly again refer you to the COC. I am more than happy to accept bug reports from you, or even better, see you actually helping the open source world and contributing PRs, which would be awesome! After all this is open source, and we, the whole community and I really mean no single other open source project of the world is due anyone anthing. Since I do not waste time on trolling or toxicity too much, this will not receive any subsequent answer, since the topic is cleared up and intentions are clear since the only thing you still do is brag around, even though I started an actual dialogue about what could be improved. This behaviour is not welcome and is not tolerated, you're still welcome to participate, but in a way that is not just profiling yourself and in a way that is constructive. Thanks for understanding. |
* cs fixes * add concepting to comments * fix driver inclusion * remove sqlite dependency, add sqlite to devdependencies and outsource driver * remove unnecessary files * remove mongodb dependency, outsource mongodb driver * Bump sqlite3 package version to fix issue with node 4.0.0 * Bump sqlite3 package version to fix issue with node 4.0.0 * add node version 4.0, remove iojs allowed build failures * remove mysql dependency and add to dev for testing purposes * fix index_test tests * change base test to extern base * fix index test * remove last driver dependencies resolves db-migrate#227 * add pg to dev dependencies * ? * updated fixed pg driver * add error message to executeDB and handle callback * return error message if available or the object/message directly passed through * fix some cs issues * add missing nosql methods to interface * do not strip private functions from driver removed unnecessary entries from migrator interface do not strip any `_function` unless it is explicitly defined via exclude * change config to internal scope * fix test for new internal config * fix last details to move config i into internal scope * remove debug message * remove es5 from jshint rules, fix one cs * add rules to jshint * fix some cs issues and adjust rules * adjust some settings and add jscs already * [cs] fixed some issues and do some cleanup * update lab * update dependency and check * add 4.1 to travis ci * change couple of templates * update sql templates * add undo seed function to api * fix bug on up/down of seeders * add reset method and limit down methods of seeds * add dummy calls to linked seeders and migrations * link to read the docs and remove readme documentation * 0.10.0-beta.1 * remove the badge ending from the documentation url * make entry in readme more clear to be also a user manual * Extend startup with logging and failure handling on 0.9.x versions Before this patch users may ran into a local installation older than 0.9.x, this patch ensures that the version is compatible to be preffered over the global version. If the local version is to old (<0.10.0) the global version will be used instead. If this is the case a warn message gets displayed to the user. Additionally added informational logging within the verbose mode if a local version gets preferred. solves db-migrate#317 * 0.10.0-beta.2 * add lts node.js version to travis tests * Resolve driver from local projects Fixes db-migrate#318 * use synchronized variant of resolve for drivers * 0.10.0-beta.3 * fix sql templates to work with 0.10.x again * ignore on init formatting * 0.10.0-beta.4 * Support setting default environment via ENV variable. * Support setting default environment via ENV variable. * coffeescript template refactory coffeescript template refactory by default coffeescript always returns the last line/variable, so there is no need to use the ~return~ keyword, there is not need for ; also. * add scope logging * [bugfix] internal linking causes infinite loop fixes db-migrate#329 * add scope logging * [bugfix] internal linking causes infinite loop fixes db-migrate#329 * coffeescript template refactory coffeescript template refactory by default coffeescript always returns the last line/variable, so there is no need to use the ~return~ keyword, there is not need for ; also. * 0.10.0-beta.5 * remove debug message * 0.10.0-beta.6 * allow configuration through options object * adjust test * add direct execution of seeders and migrator links * Fix config file setting. * update parse db-url to 0.3.0 Introduces mongodb uri support * update some dependencies * fix(migration): A promise is not properly identified and returning an empty resolve ends up in db-migrate hanging up forever. Fixes db-migrate#343 * fix(api): callback not called on run If db-mgirate is run via any run method (except directly from console), the callback was not available to the migration functions anymore. * fix(api): callback called twice Revert the change to add the callback if it is undefined. Fixes db-migrate#343 * Fix promise issue for down migrations. * refactor(migration): Make check for promise more reliable * fix(test): Stub MySQL connect method instead of calling the original The MySQL driver now connects directly to the driver, instead on the first query. This caused some trouble with the ssh tunnel test, this functionality was stubbed. Fixes db-migrate#348 * chore(security): Bump dependency because of security vulnerability * pull down package.json * chore(dependencies): update and remove unneeded dependencies * Update data_type.js For consistency :-) * chore(travis): change travis config to use modern compiler * chore(travis): add node 5 to ci builds * chore(travis): remove obsolete node 4.1 Keep 4.2 lts and node 4 in the build which is usually enough. * Avoid warning in promise/callback bridge code When Bluebird detects when a promise is created within a promise handler, but is not returned, it [emits a warning][warn] 'a promise was created in a handler but was not returned from it'. This happens if the `callback()` function is implemented using Promises, but uses `nodeify()` to make it callback compatible. By using nodeify here, it handles the returned Promise (if any) and avoids the warning. [warn]: https://github.com/petkaantonov/bluebird/blob/master/docs/docs/warning-explanations.md * Updated documentation link in README Current documentation link points to a 404 page, updated it with a working version. * chore(travis): add node 6 * fix(api): fix race condition on create migration A callback was placed not inside but after the callback which it belongs to. This resulted in a very rarely occuring race condition when creating migrations without an existing migration directory. Fixes db-migrate#376 * 0.10.0-beta.12 * chore(compatibility): update dependencies * chore(dev): add .tern file to project * refactor(log): move log into new extern lib Refers to db-migrate#382 * refactor(datatype): move dataType into new extern lib Refers to db-migrate#382 * refactor(remove): remove log and dataType artifacts Refers to db-migrate#382 * refactor(util): remove util artifact Resolves db-migrate#382 * detect real callback for several api functions * feat(api): promisify all current api methods * refactor(ssh): delay loading of tunnel-ssh and improve load time This improves the loadtime of db-migrate by 47%. To be exact from ~22ms to ~15ms. Resolves db-migrate#387 * refactor(modularize): require inflection from npm instead * refactor(tests): added shadow test and rewrite api test * chore(comments): add some notes to the api tests * refactor(globals): remove several globals, this might break things There are several things that are exposed as global. For example Migrator, Seeder, Class, or the async and dbm globals. Most probably finally removing async and dbm which have been marked deprecated, will end up for some people in a breaking change. But we need to remove all those globals, to not interfere with projects using db-migrate. DB-Migrate is not just a cli module anymore though. In this step we now also finally removed internals completely from the API, and needed to adjust the onComplete callback to actually do this. This change could also lead to a breaking change and should thus be communicated All in all this is a big clean up and also two not necessary functions were removed. * chore(parallelize): parallelize non interferring api tests * refactor(tests): add index tests and some configs * refactor(tests): refactored util_tests * refactor(tests): refactored migration tests * refactor(tests): refactored config tests * refactor(tests): refactored base tests * refactor(tests): add create tests and remove sample shadows * chore(ci,dep): update ci and dependencies * chore(test): create Makefile for tests and edit ci config Removing 0.10 and 0.12 from tests * chore(ci): clean up ci config * chore(dependencies): upgrade dotenv and pkginfo * chore(devdep): update and remove some devDependencies * use path.join (db-migrate#353) refactor(template): use path.join as it is intended to be used * fix(template): fix unnoticed error introduced in the last merge request Wrong migrations have been generated * feat(plugins): add basic support for plugins and improve performance Some libraries have been changed to be delay loaded. DB-Migrate is back to the minimal load time possible again, if no command at all is entered. Basic plugin support and the first hook, in this case of the type overwrite has been added. An functional example of a yaml config plugin has been published. Also some old functions have been cleaned up which are not necessary any more. refers to db-migrate#397 * chore(leftover): remove example plugin leftover from package.json * fix(api): fix introduced undefined behavior of specified configs internal.cwd added again, which is needed. * fix(api): fix scoping The api was missing the matching internal, which resulted in a missing prefix when executing scope migrations. fixes db-migrate#409 * most of the first protocol transition and the transition helper is implemented already. Just the logic to actually call the helper is currently missing. * feat(transitioner): add transitioner to easen the process of protocol changes Resolves db-migrate#403 * feat(hook): parser hook and transitioner api Added a hook for parsers and added the API endpoints for the transitioner. Refers to db-migrate#403 Refers to db-migrate#397 * fix(config): Don't throw if environment variable is empty Fixes db-migrate#411 * chore(hook): allow also to hook without adding any extension * fix(transitioner): add new parser internal to transitioner * fix(transitioner): catch whitespaces properly * chore(templates): add meta protocol v1 to already aligned migration templates * feat(plugin): add basic plugin support Added basic plugin support, enabling to inject at multiple targets of the API, config and register pre compiler. Resolves db-migrate#397 Refers db-migrate#396 * fix(plugin): use correct path to include plugins * fix(errorhandling): Add missing error assertion in executeDB fixes db-migrate#381 * feat(sync): add sync mode Refers to db-migrate#383 Resolves db-migrate#313 Refers to db-migrate#222 * refactor(tests): remove tests from main repo Tests have been moved to their driver repo. Resolves db-migrate#399 * refactor(cleanup): cleanup package.json and add db.config.ci back * chore(ci): exclude interfaces * chore(ci): Readd missing dependencies * chore(test): Add Test for privateKey param on shh tunnel * fix(resolve): Check if resolved version has plugin support fix db-migrate#425 * log error on undefined stack * refactor(api): cleanup information and added commentary * feat(config): add rc style configs fixes db-migrate#308 fixes db-migrate#406 * refactor(api): refactor tests and make api.js more modular * fix(api): add missing reference to sync * refactor(api): remove last output of postponed seeders * add test dummies for api * add first programable api tests * directly use promises in tests * Double quoting interpolated file name for coffee * fix(create): use same timestamp in every created file * fix(args): dont parse when called as module fixes db-migrate#449 * fix deep config replacements fixes db-migrate#473 * feat(config): helper to overwrite and extend configuration This allows to parse a url from an url property and as well to overwrite and extend variables as needed. To pass in url as normal value an object is needed like the following: { url: { value: "http://example.com } }. Fixes db-migrate#349 Fixes db-migrate/pg#8 Fixes db-migrate#488 Fixes db-migrate#463 * fix(tests): fix breaking tests * refactor(config): do not just interpret database.json also check for either a passed config object OR a config file exists * fix(create): Fix create when using db-migrate as module Bug introduced on db-migrate#485 Fixes: db-migrate#493 * test(create): Fix testing as module when no database.json on root * fix small typo in cli help menu * allow rc file to contain location of config file * print the reason in unhandledRejection evnet if reason.stack does not exist * complete refactor of api js * add eslint to dependencies * fix a few bugs and update tests * fix last tests * adjust to cs * adjust manual changes for cs * remove process.exit when no callback passed and remove optimist help when in module mode Fixes db-migrate#516 * add changelog and release 0.10.0 * 0.10.0 * adjusted tests and updated devDependencies Refers to db-migrate#518 * readd pkg-lock * change tested versions * rewrite latest to node * add linting and fix cs on config test * add commitlint * remove parallel: true as it was deprecated * add stale config * use more appropiate label * Create CODE_OF_CONDUCT.md * fix(db): wrong reference to connect causes db:create to fail Fixes db-migrate#520 * 0.10.1 * add changelog * fix(log): error ended up in unreadable errors Fixes db-migrate#524 Fixes db-migrate#521 * add changelog * chore: Remove `moment`. Signed-off-by: wtgtybhertgeghgtwtg <wtgtybhertgeghgtwtg@gmail.com> * feat(issuetemplate): added a github issue template Signed-off-by: BorntraegerMarc <marc.borntraeger@gmail.com> * fix(exitCode): wrong check for existence fixed A wrong check for existence resulted in events not being loaded on version sized like 4.0.0. This resulted in non zero error codes when actually running into errors. Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * refactor(deprecationNotice): remove insert deprecation This temporarily removes the note being displayed to the user when an insert call is being made. This will be removed until seeders really going to be introduced. Refers to db-migrate#215 Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * fix(exitCode): wrong exit code on db methods DB Calls always returned an exit code of 0, which leads to an unexpected behavior on the user side. Fixes db-migrate#534 Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * fix(switchDatabase): no error was thrown on scope switch switchDatabase error was unhandled, which resulted in an unhandled scope switch error. Fixes db-migrate#470 Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * 0.10.3 * add changelog * fix(insert): add missing insert entry to interface Fixes db-migrate#542 Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * 0.10.4 * changelog Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * 0.10.5 * update to use real checkboxes * Update ISSUE_TEMPLATE.md * feat(contribution): enrich contribution instructions Refers to db-migrate#549 Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * feat(contribution): enrich contribution instructions, issues Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * feat(progamableApi): CMD options can be passed programatically now Signed-off-by: Marc Bornträger <marc.borntraeger@gmail.com> * fix(ci): add ignores for backported features Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * chore: update dependencies Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * feat(progamableApi): using const now Signed-off-by: BorntraegerMarc <marc.borntraeger@gmail.com> * add changelog Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * 0.10.6 * chore: add test to detect cmdOptions functionality Refers to db-migrate#560 Refers to db-migrate#557 Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * fix(progamableApi): cmdOptions get passed into setDefaultArgv now Signed-off-by: Marc Bornträger <marc.borntraeger@gmail.com> * 0.10.7 * add changelog Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * chore: update dependency Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * chore: update dependency Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * feat(check): add check functionality to determine migrations to run Signed-off-by: RandomSeeded <nate@blend.com> * 0.11.0 * add changelog Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * fix(reset): regression introduced in check functionality db-migrate#552 introduced a regression and effectively disfunctioned the reset function, since it did append just to the actionables within the run context. It was moved into its own context. Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * add changelog Signed-off-by: Tobias Gurtzick <magic@wizardtales.com> * 0.11.1 * fix(check): fix check via API not passing results to the callback Signed-off-by: RandomSeeded <nate@blend.com> * save * save * updates * cleanup
Disclaimer: I may be totally off base on what's actually happening here, if I am please let me know 😅
Overview
The API shouldn't quit the process, it should throw an error so that clients can catch the issue.
Issue
When the API fails, it sometimes throws a
process.exit(1)
which can lead testing frameworks (Jest) to misplace blame. Since Jest (and Ava) run tests in parallel,process.exit(1)
can stop the whole shebang in the middle of something leading to weird errors like this:For Screenreaders:
In the example above, a test that hadn't reported in yet may have called the
process.exit(1)
, but since that interrupted the output ofsrc/lib/core/test/core.test.js
, it appears as if that test is the cause.At the moment, it looks like the
api.js
node-db-migrate is calling the commands directly viaload
.Take the
create
command, as reproduced below:That will call the
lib/commands/create-migration.js
file, which callsprocess.exit(1)
in several places. For example, if it fails to create a migration:Solution
Whenever we would call
process.exit(1)
, check if we're running in the API and instead throw an error? Or maybe redefineprocess.exit
to something like this:If this is okay, I would be willing to take a stab at it. If not, no worries.
The text was updated successfully, but these errors were encountered: