Skip to content

fix(thread): a translating message shows placeholder bars down its whole height - #5141

Closed
stamp-build[bot] wants to merge 1 commit into
mainfrom
agent/lagoon/translation-skeleton-lines
Closed

stamp-build[bot] wants to merge 1 commit into
mainfrom
agent/lagoon/translation-skeleton-lines

Conversation

@stamp-build

@stamp-build stamp-build Bot commented Nov 20, 2026

Copy link
Copy Markdown
Contributor

Lagoon from raft opened this pull request through Stamp
View in Stamp →


artin (DM, 2026-11-20): the translation keeps its height well, but the skeleton has only the first line.

Since #5103 (#201), a message with a pending automatic translation keeps the original body's height, but the placeholder was only the web's single 176dp bar at the top of that space. Now ThreadMessageTranslationPendingFrame draws one 12dp bar per line of the original (8dp gap, 20dp pitch, at most 16 bars). The bars are full width and the last one is 176dp. A one-line message looks the same as before. Only the bars that get placed are measured. The original is still measured only, never placed, and the row height is unchanged.

Tests

  • New aPendingTranslationDrawsOneBarPerLineOfTheOriginal (line-count table).
  • The existing aPendingTranslationKeepsTheOriginalsHeightWithoutShowingIt still guards the height.
  • CI runs both.

No changeset: #5103 hasn't shipped yet (it's in 1.27.0). This will be cherry-picked to release/1.27.0.

🤖 Generated with Claude Code

…ole height

#5103 keeps a translating row at the original body's height, but drew
only the web's single 176dp bar at the top, so a tall message looked
empty. The placeholder now draws one 12dp bar per line of the original
(8dp apart, up to 16), full width with a 176dp last line.

Co-authored-by: bytemain <13938334+bytemain@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@stamp-build

stamp-build Bot commented Nov 20, 2026

Copy link
Copy Markdown
Contributor Author

Merged by Stamp in batch 6a2458c9389667280f1501268968583f: c78a6a1 . The batch passed its required CI; this PR is closed because its commit was included by fast-forward.

Review in stamp.build

@stamp-build stamp-build Bot closed this Nov 20, 2026
stamp-build Bot added a commit that referenced this pull request Nov 20, 2026
…sion, Assistant row, translation placeholder, reset-notice polling) (#5158)

* fix(auth): Home re-reads the two-step reset notice at most every 30 s (#5137)

Stamp-Batch: 6a2458c9389667280f1501268968583f
Original-Head: 84a763f
Signed-off-by: Stamp Queue <stamp-queue@stamp.build>
(cherry picked from commit 636993f)

* fix(thread): a translating message shows placeholder bars down its whole height (#5141)

Stamp-Batch: 6a2458c9389667280f1501268968583f
Original-Head: 32fe17d
Signed-off-by: Stamp Queue <stamp-queue@stamp.build>
(cherry picked from commit c78a6a1)

* fix(thread): composer uploads key on the signed-in account from the first frame (#5143)

Stamp-Batch: 7fa83bbd6a5680b3d5cdd4a2621e7d0d
Original-Head: 53a7a74
Signed-off-by: Stamp Queue <stamp-queue@stamp.build>
(cherry picked from commit 67e5dfc)

* fix(assistant): phone Home shows the Assistant row (the layout decides, not the desktop-mode setting) (#5151)

Backport of #5151 (head cde2b9f) to release/1.27.0.

* fix(thread): an edited or annotated image no longer sticks on "Checking file…"

The event-bus effects that take the image editor's Done (and the other
composer events) are keyed only on the thread, so they kept the first
frame's copy of uploadComposerAttachmentFiles. On that frame the session
is not loaded yet, so the copy holds the upload store of account "".
The upload ran and succeeded there, but the composer applies outcomes
from the signed-in account's store, so the tile never left "Checking
file…". Since #5104 a cached conversation renders the full page on its
first frame, before the session is captured, which exposed this on
Android 1.27.0.

The effects now call the latest copy through rememberUpdatedState.

* fix(thread): a share sent while its files upload is sent with the loaded session

The external-share auto-send collector is keyed only on the thread, so it
kept the first frame's performComposerSend, whose session and provenance
are null before the session loads: the send was rejected
(authority_changed_before_message_send) and the share never went out.
It now calls the latest copy through rememberUpdatedState.

* fix(thread): latest-upload and latest-send wrappers are lambdas, not function references

A local function reference compares equal across compositions (equals
ignores the captured locals), so rememberUpdatedState's structural
policy kept the first frame's copy and the previous commit had no
effect. Found by Reef in the alpha's dex.

* fix(thread): seed first-frame session and provenance on cached opens and harden null-owner effects

* refactor(thread): extract performance-gate and task-conversation helpers from SharedThreadRouteContentContent

Co-authored-by: Lagoon <raft+lagoon@agents.stamp.build>
Co-authored-by: Reef <raft+reef@agents.stamp.build>
Co-authored-by: bytemain <13938334+bytemain@users.noreply.github.com>
stamp-build Bot added a commit that referenced this pull request Nov 21, 2026
…re already on main) (#5189)

* docs(release): 1.27.0 Android, iOS and Google Play notes

Aggregated from the 126 accepted changesets at release/1.27.0 (9c9f114).
Diagnostics-only changesets omitted; flagged rollouts marked as gradual.

* docs(release): 1.27.0 notes leave out the personal assistant (still gated)

* fix: 1.27.0 release candidate fixes (stuck upload and first-frame session, Assistant row, translation placeholder, reset-notice polling) (#5158)

* fix(auth): Home re-reads the two-step reset notice at most every 30 s (#5137)

Stamp-Batch: 6a2458c9389667280f1501268968583f
Original-Head: 84a763f
Signed-off-by: Stamp Queue <stamp-queue@stamp.build>
(cherry picked from commit 636993f)

* fix(thread): a translating message shows placeholder bars down its whole height (#5141)

Stamp-Batch: 6a2458c9389667280f1501268968583f
Original-Head: 32fe17d
Signed-off-by: Stamp Queue <stamp-queue@stamp.build>
(cherry picked from commit c78a6a1)

* fix(thread): composer uploads key on the signed-in account from the first frame (#5143)

Stamp-Batch: 7fa83bbd6a5680b3d5cdd4a2621e7d0d
Original-Head: 53a7a74
Signed-off-by: Stamp Queue <stamp-queue@stamp.build>
(cherry picked from commit 67e5dfc)

* fix(assistant): phone Home shows the Assistant row (the layout decides, not the desktop-mode setting) (#5151)

Backport of #5151 (head cde2b9f) to release/1.27.0.

* fix(thread): an edited or annotated image no longer sticks on "Checking file…"

The event-bus effects that take the image editor's Done (and the other
composer events) are keyed only on the thread, so they kept the first
frame's copy of uploadComposerAttachmentFiles. On that frame the session
is not loaded yet, so the copy holds the upload store of account "".
The upload ran and succeeded there, but the composer applies outcomes
from the signed-in account's store, so the tile never left "Checking
file…". Since #5104 a cached conversation renders the full page on its
first frame, before the session is captured, which exposed this on
Android 1.27.0.

The effects now call the latest copy through rememberUpdatedState.

* fix(thread): a share sent while its files upload is sent with the loaded session

The external-share auto-send collector is keyed only on the thread, so it
kept the first frame's performComposerSend, whose session and provenance
are null before the session loads: the send was rejected
(authority_changed_before_message_send) and the share never went out.
It now calls the latest copy through rememberUpdatedState.

* fix(thread): latest-upload and latest-send wrappers are lambdas, not function references

A local function reference compares equal across compositions (equals
ignores the captured locals), so rememberUpdatedState's structural
policy kept the first frame's copy and the previous commit had no
effect. Found by Reef in the alpha's dex.

* fix(thread): seed first-frame session and provenance on cached opens and harden null-owner effects

* refactor(thread): extract performance-gate and task-conversation helpers from SharedThreadRouteContentContent

* chore(release): merge back 1.27.0 (release notes; the release fixes are already on main)

Squash of the merge of v1.27.0 into main. What main gains: the 1.27.0 Android, iOS and Google
Play notes (change-logs/1.27.0/{android,ios,play}). The release branch's code fixes (#5137, #5139,
#5141, #5143, #5151, squashed there as #5158) were cherry-picked from main, so main already has
them; ThreadPage.kt keeps main's newer version.

Co-authored-by: stamp-build[bot] <328845999+stamp-build[bot]@users.noreply.github.com>
Co-authored-by: Lagoon <raft+lagoon@agents.stamp.build>
Co-authored-by: Reef <raft+reef@agents.stamp.build>
Co-authored-by: bytemain <13938334+bytemain@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants