- 11 Feb, 2025 16 commits
-
-
vercel-release-bot authored
-
Jiachi Liu authored
-
Jiachi Liu authored
-
Tobias Koppers authored
### What? avoid the extra work when increasing active count
-
Tobias Koppers authored
### What? * allow to disable active tracking * disable active tracking when dependency tracking is disabled This allows to skip a lot of active counter increases and decreases for better performance. Without dependency tracking there can't be any inactive tasks in the system, so we don't need to track that at all.
-
Niklas Mischkulnig authored
-
Tobias Koppers authored
### What? Make node.js process selection more sticky to the worker where we spend the most time operating on.
-
Tobias Koppers authored
### What? fix cases where action counter is messed up when tasks are invalidated during execution
-
Tobias Koppers authored
### What? We don't need to duplicate static modules per asset context. They are independent of it.
-
Tobias Koppers authored
### What? Add an option for using the parser AST state and only enable it in development mode.
-
Tobias Koppers authored
### What? * add tracing to recomputation * fix recomputation hanging in a loop (reexecution should not invalidate recomputed cells, even if they still have content and Eq says they are not equal. Some value types does implement Eq as always false, e. g. the AST)
-
Benjamin Woodruff authored
Extracted from the try-ci-test work: #75443 Includes some minor cleanup, deduping some use of the environment variable. <!-- Co-authored-by:JJ Kasper <jj@jjsweb.site>
-
Benjamin Woodruff authored
Extracted from the try-ci-test work: #75443 I'm just trying to find small unobjectionable pieces we can get merged to get the size of the in-flight changeset down. <!-- Co-authored-by:JJ Kasper <jj@jjsweb.site>
-
vercel-release-bot authored
-
Jiachi Liu authored
### What We should not throw the postpone error in metadata resolving when streaming metadata is not enabled. ### Why The suspense boundary of streaming metadata can catch the postpone error and render the fallback, but in blocking metadata world we should just swallow it otherwise it will send empty fallback shell
-
Benjamin Woodruff authored
- Removes all the branches between Turbopack and webpack (depends on https://github.com/vercel/next.js/pull/75730) - Makes the test less dependent on log ordering (which isn't really relevant or important here) which should also unbreak it for Rspack. - Reduces the amount of code inside the `retry` block. - Update comments.
-
- 10 Feb, 2025 17 commits
-
-
vercel-release-bot authored
-
Jiachi Liu authored
-
Hendrik Liebau authored
-
Hendrik Liebau authored
The Resume Data Cache (RDC) is used to provide cache consistency when resuming a partial prerendered static shell. When reading from this cache, it is not possible for there to be a miss because, regardless of the cache age, we always consume the cache entries on resume because they are temporarily bound to the static shell being served for this particular request. Since the cache entry can never miss, there is no need to store inner cache entries in the Resume Data Cache because the only time you would encounter an inner entry is if the outer entry had to revalidate. We have already optimized the Resume Data Cache by excluding writing to it during prerendering (and dev warmup rendering) if we're already inside a cache scope. We also handle that you can have an inner cache entry fill the backing cache and omit the RDC entry, and then later we encounter that same cache entry in the prerender scope, and even though we don't need to produce a new cache entry (it was already cached in the backing cache implementation), we do still write it to the Resume Data Cache. The test added in this PR verifies this behavior and ensures that we don't regress in the future. closes NAR-54
-
Benjamin Woodruff authored
Previously, when making a change that doesn't actually cause any modules to update, you could see a: ``` [Fast Refresh] rebuilding ``` that is never followed by a ``` [Fast Refresh] done in 90ms ``` message. This PR attempts to more closely emulate webpack's behavior. I've seen people observe this before and wrongly believe that Turbopack is hanging. Doing this right is a little tricky: Turbopack emits the `BUILT` event separately using a debounce (currently 30ms) after the actual HMR updates. I don't think the Turbopack HMR APIs are particularly great overall, but the way Turbopack is handing this sort-of makes sense. A file could be updated multiple times in quick succession (i.e. as the editor is in the middle of writing it), or somebody could save changes to multiple files at once, and there's no way to atomically synchronize all of the potential filesystem/editor interactions.
-
Sebastian "Sebbie" Silbermann authored
Failing test highlighting breakage in https://github.com/vercel/next.js/pull/75294
-
Jude Gao authored
-
Jiwon Choi authored
Following up of #75679 ### Why? Clicking the navigation escapes the overlay. It is because the navigation component is outside of the dialog component. #### Before https://github.com/user-attachments/assets/e7c63261-c44b-4af7-936d-1750dff07527 #### After https://github.com/user-attachments/assets/979421fa-fb71-481d-bd58-6703298db28b ### How? Since we already have logic to prevent closing overlay when clicked outside the component for dev indicator, add the header to this CSS selectors list as well.
-
Vercel Release Bot authored
-
Jude Gao authored
This configure step is potentially required for the ports to show up in the remote targets.
-
rauno authored
This PR implements a new design for the experimental Next.js Error Overlay: - Border-less body design - Notch cut-out for the navigation and version info <img width="1088" alt="image" src="https://github.com/user-attachments/assets/31385793-82c4-41ab-a34d-6b49955f5823" /> --- - Closes [NDX-782](https://linear.app/vercel/issue/NDX-782/re-redesign-contents-in-a-single-body-with-padding) - Closes [NDX-781](https://linear.app/vercel/issue/NDX-781/re-redesign-navigation-header) --------- Co-authored-by:
devjiwonchoi <devjiwonchoi@gmail.com>
-
rauno authored
### Why? The bottom stack style was regressed.  This PR closes [NDX-808](https://linear.app/vercel/issue/NDX-808/fix-regression-in-bottom-stacks-style) --------- Co-authored-by:
Jiwon Choi <devjiwonchoi@gmail.com>
-
-
vercel-release-bot authored
-
Jiachi Liu authored
Reverts vercel/next.js#75366 Noticed there's bug with `__nextppronly=fallback` even streaming metadata is not enabled in PPR
-
Sebastian "Sebbie" Silbermann authored
-
Vercel Release Bot authored
Co-authored-by:Sebastian Sebbie Silbermann <sebastian.silbermann@vercel.com>
-
- 09 Feb, 2025 5 commits
-
-
vercel-release-bot authored
-
Jiachi Liu authored
### What missing html tags error should always show the overlay when exist, and the dev indicator count should be 1 since we require users to resolve this first then React can hydrate the right content. After this they can move to others. Since the indicator count is not covered in old test, I add a new test for new overlay rn, later we can merge them #### Observed * Currently it requires to click the indicator to show the error * The indicator count shows 2 #### Expected * The overlay should pop up as it's a runtime error * The indicator count shows 1
-
Lee Robinson authored
Co-authored-by:Jiachi Liu <inbox@huozhi.im>
-
Jiachi Liu authored
### What `[data-nextjs-dialog-content]` was set to `height: 100%`, which has conflicted with useMeasureHeight hook. `data-nextjs-dialog-content]` 's parent is always controlled by `useMeasureHeight`, and the height is anmiating from 0 to the child height. If the child is aligned with that parent, then the height is always 0, which lead to not displayed Closes NDX-797
-
Sebastian "Sebbie" Silbermann authored
-
- 08 Feb, 2025 2 commits
-
-
vercel-release-bot authored
-
Jiachi Liu authored
-