* Add a check for existing .git directory (and skip if one is found).
* Changeset attempt :-)
* Update .changeset/try-button-rumor.md
Co-authored-by: Nate Moore <natemoo-re@users.noreply.github.com>
* `create-astro`: always create `tsconfig.json`
Currently, we only make sure `tsconfig.json` exists when `strict` or `strictest` is selected. Both `default` & `optout` are intended to correspond to `base` -- and will do so for all [23 official templates](https://github.com/withastro/astro/tree/main/examples), but not necessarily for third-party templates.
The [example command for installing a third-party template](https://github.com/withastro/astro/blob/a800bf7/packages/create-astro/README.md?plain=1#L31-L35) is (rather conveniently for the sake of this PR!) an example of a template without a `tsconfig.json` file, and installing it with the `default` ("Relaxed") Typescript option results in no `tsconfig.json` file, rather than a `tsconfig.json` file containing `{ "extends": "astro/tsconfigs/base" }` as would be expected.
This PR addresses this scenario.
It also explicitly sets the `tsconfig.json` file to `{ "extends": "astro/tsconfigs/base" }` when `default` (which I renamed to `base`, still presented to the user as "Relaxed") or `optout` is selected (`optout` has always printed a warning about the importance of `tsconfig.json` & `src/env.d.ts` but otherwise behaved identically to `default`). This is necessary in two scenarios:
1. When the `tsconfig.json` file was created by this script.
2. When it either didn't already include `"extends"`, or it extended a different config by default. For example, some third-party templates might default to `strict`, in which case I'm guessing we'd want to respect the user's choice and change that to `base`.
* update `del` 6.1.1 --> 7.0.0
* test: prevent excess writes
(without this it triggers many times)
* test: create-astro typescript prompt
* changeset
* fix: recursive `mkdirSync`
* test: longer timeout for `windows-latest` OS
(see if this fixes failing tests)
* better glob path creation, don't hardcode `/`
* test: longer timeout for windows-latest OS
(since I'm about to trigger another CI run by pushing a commit, might as well try this too)
* create-astro test: show last CLI output on timeout
* drop variable timeout
Typescript tests are slower than directory tests, but they are all usually less than 5000 ms. Less complexity, easier to maintain.
* DRY new error output
* Update lockfile
* Sync lockfile with main
* Update lockfile
Co-authored-by: Princesseuh <princssdev@gmail.com>
* enabling eslint on the all packages and tests
* enabling for all packages
* TEMP: adding an only() test to verify it fails CI
* using our eslint config and ignore in CI
* removing the temporary .only() test
* update lock file
* lint: fixing new test with a no-shadow warning
* chore: update lock file
* fix(#3309): use system default locale
* fix(#3309): use system default locale in create-astro
* test: add locale regression tests
* test: add i18n regression test
* feat: add git init step
* fix: update unit tests
* feat: simplify next steps for copy pasteability
* docs: add clarifying comment on test stdin spoofing
* docs: remove "empty" from git repo message
* fix: update git step text for test
* fix: remove redundant --dryrun flag
* refactor: simplify next steps with &&
* chore: changeset
* refactor: simplify dir error log to avoid wrapped text
* refactor: remove redundant "issue" callout
* chore: changeset
* chore: update tests for new dir log
* wip: add prompt for directory with validation
* feat: wire up dir response to cwd
* feat: improve error handling on non-empty dirs
* fix: update test helpers to execaSync
* chore: add .skipped to old tests for clarity
* deps: add mocha and chai to create-astro
* feat: add directory step test with fixture
* feat: update turbo to run create-astro tests again 🥳
* chore: changeset
* update examples
* add initial integrations
* update tests
* update astro
* update ci
* get final tests working
* update injectelement todo
* update ben code review
* respond to final code review feedback
