2004961 - Stale unread count may persist in folder ARIA label/tooltip after messages are read or deleted

Status: RESOLVED FIXED

(Thunderbird :: Disability Access, defect)

Description

[Colomban Wendling]

Attachment: tb-unread-count-a11y-bug.png

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:140.0) Gecko/20100101 Firefox/140.0

Steps to reproduce:

Delete all unread messages in a subfolder (e.g. INBOX/subfolder)

Actual results:

The unread count badge for that directory in the accounts list properly updated (the badge disappeared as there are no more unread messages) but the a11y description of the row kept the previous value ("INBOX/subfolder, 42 unread messages" -- or so, reverse-translating from French).

Expected results:

The a11y description should update properly, even when dropping to no unread messages, and even in subfolders.

---

In the attached capture, you can see that the a11y treeitem node for folder Courrier entrant/debian (translating to Inbox/debian) still has a previous name value, while it's properly updated in the UI (hidden badge)

I am not entirely sure if this has to happen in a subfolder or if it's partly random, but I've been experiencing it there. I just tested with Daily 148.0a1 (2025-12-09) (64 bits) under Debian testing, but it's been like that for months at least.

Comment 1

[Magnus Melin [:mkmelin]]

Right, looks like aria-label isn't updated for 0 unread, leaving the old value there.
https://searchfox.org/comm-central/rev/b3656e2549ad7bf5b5c167eb5dc538c2fbf29448/mail/base/content/widgets/folder-tree-row.mjs#208

Comment 2

[Colomban Wendling]

@mkmelin thanks for looking into this!

Without knowing this code though, are you sure that's where lies the issue? It looks to me like the aria label is always updated with the promise pieces (L227), and that L208 just avoids adding the unread label to the promise.

At a glance, I'd be wondering whether this.readCount could still be at an old value when #updateAriaLabel() is fetching it? Or if there could be promise ordering issues as all of unreadCount, totalCount and folderSize will be updating it pretty much all at the same time? I hope not because that sounds awful to debug :)

I hope I'm wrong and just missed something though :)

Comment 3

[Magnus Melin [:mkmelin]]

I didn't try coding it so there is a chance my analysis is wrong.

Comment 4

[Hartmut Welpmann [:welpy-cw]]

While the visible row in the folder pane is changed immediately (https://searchfox.org/comm-central/rev/7f4bed3d3e3105bc10a41a0fa7c0327a21f6f59c/mail/base/content/widgets/folder-tree-row.mjs#150-161), the aria label is localized asynchronously (https://searchfox.org/comm-central/rev/7f4bed3d3e3105bc10a41a0fa7c0327a21f6f59c/mail/base/content/widgets/folder-tree-row.mjs#227-234).

You say this happened when you deleted all unread messages, so maybe automatic compacting of the folder might kick in right afterwards, delaying the fluent localization for the aria label. Does the label eventually get updated on its own? By the way, you can easily check this by hovering the mouse and having the tooltip displayed.

Comment 5

[Wayne Mery (:wsmwk)]

Does the label eventually get updated on its own? By the way, you can easily check this by hovering the mouse and having the tooltip displayed.

Comment 6

[Colomban Wendling]

You say this happened when you deleted all unread messages, so maybe automatic compacting of the folder might kick in right afterwards, delaying the fluent localization for the aria label. Does the label eventually get updated on its own?

It doesn't seem so no, it also happens removing a single email (although it's the only unread), and waiting a few minutes doesn't seem to help. Also, there's no visible mention of compacting going on.

By the way, you can easily check this by hovering the mouse and having the tooltip displayed.

Ah yeah good point, the tooltips are wrong as well, still include how many unread there was before deletion.

Comment 7

[Colomban Wendling]

Attachment: tb-noread-1.png

Showing 1 shadow unread mail on the Inbox (in French)

Comment 8

[Colomban Wendling]

Attachment: tb-noread2.png

Showing 13 shadow unread mails on a subfolder (in French)

Comment 9

[Hartmut Welpmann [:welpy-cw]]

Attachment: Bug 2004961 - Refactor FolderTreeRow state management and fix ARIA label race condition. r=darktrojan,#thunderbird-front-end-reviewers

This commit addresses several performance and state management issues within the FolderTreeRow component:

• Decoupled State: Migrated unreadCount, totalCount, and folderSize from
  reading directly from the DOM (textContent) to using private class fields
  as the single source of truth. Added early-exit equality checks to setters to
  prevent unnecessary DOM work.
• Race Condition Fix: Implemented an AbortController inside #updateAriaLabel
  to ensure older, delayed localization promises do not overwrite newer state
  updates.
• Performance/Debouncing: Added #scheduleAriaLabelUpdate to batch rapid,
  sequential property updates (e.g., during an email sync) into a single 50ms
  delayed localization call.
• Layout Optimization: Centralized the nodeValue update logic into
  #updateTextNode to successfully bypass the childList MutationObserver
  across all folder labels, preventing unnecessary layout recalculations.
• Documentation: Corrected JSDoc for folderSize to accurately reflect its
  string return type.

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/79cfa268c280
Refactor FolderTreeRow state management and fix ARIA label race condition. r=darktrojan,aleca

Comment 11

[John Bieling]

We had to back-out this patch, because it introduced intermittent test fails for mac:
https://treeherder.mozilla.org/jobs?repo=try-comm-central&revision=631b14e88444c4bec341531356be100ba9eb0fbb&group_state=expanded&selectedTaskRun=PTGbLsOWTQqYihi3wAwmHA.0

We have run a try without it and could not trigger the same failure:
https://treeherder.mozilla.org/jobs?repo=try-comm-central&revision=756f740c90555ccfb20bb3668b65d12afed53801

Geoff did some investigations, but could not find what might be causing it.

@welpy-cw : Do you have an idea?

Comment 12

[Pulsebot]

Backout by [e-mail]:
https://hg.mozilla.org/comm-central/rev/fa300c65cf72
Backed out changeset 79cfa268c280 for potential test fails. r=me

Comment 13

[Hartmut Welpmann [:welpy-cw]]

(In reply to John Bieling from comment #11)

We had to back-out this patch, because it introduced intermittent test fails for mac:
https://treeherder.mozilla.org/jobs?repo=try-comm-central&revision=631b14e88444c4bec341531356be100ba9eb0fbb&group_state=expanded&selectedTaskRun=PTGbLsOWTQqYihi3wAwmHA.0

We have run a try without it and could not trigger the same failure:
https://treeherder.mozilla.org/jobs?repo=try-comm-central&revision=756f740c90555ccfb20bb3668b65d12afed53801

Geoff did some investigations, but could not find what might be causing it.

@welpy-cw : Do you have an idea?

Yes, I have a fix for bug 1962967 that should resolve this. But we have to wait until Treeherder is working again to be sure…

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/b72fa3cfa74b
Refactor FolderTreeRow state management and fix ARIA label race condition. r=aleca,darktrojan