nemo-places-sidebar: fetch free-space info asynchronously (#3764)

[leigh123linux]

Free-space info for Home, File System, mounted volumes, and
bookmarks was fetched with a single blocking GIO call per row from
update_places() on the main thread. A gvfs mount that never answers
(e.g. an Android phone connected via MTP with "No data transfer")
froze the whole sidebar - and with it, every open Nemo window -
indefinitely.

Rows are now added with no free-space info, and an async
g_file_query_filesystem_info_async() call fills it in later if the
backend answers before the row is gone. Bookmarks pointing into a
mount are covered the same way, since they were exposed to the same
hang; the default XDG bookmarks (Pictures, Documents, etc.) are left
alone, as they're always plain subdirectories of Home rather than
distinct mounts.

mtp:// and gphoto2:// mounts are excluded from this outright rather
than queried and relied upon to error or cancel cleanly. Testing
against a real device showed that a stuck query on these backends
doesn't actually get freed by cancellation - gdb still shows a
thread parked inside the MTP backend indefinitely - so retrying or
tracking such a request is pointless and, if repeated, risks pinning
GIO's shared worker-thread pool on a backend that's never coming
back. No thread or timeout is needed for the general case: the
async call itself never blocks the caller regardless of whether or
when a well-behaved backend responds.

[Copilot]

🟡 Changes recommended

The async callback can dereference disposed sidebar state (notably sidebar->store) if a backend completes successfully after disposal/cancellation, risking a crash.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR prevents Nemo’s places sidebar (and thus open Nemo windows) from freezing when querying filesystem free-space info for certain mounts by moving those queries off the main thread and populating the UI first, then filling in disk usage later.

Changes:

• Replace synchronous g_file_query_filesystem_info() “per row” calls with g_file_query_filesystem_info_async() and a row-reference based callback update.
• Track and cancel in-flight free-space queries via a shared GCancellable across sidebar refreshes / disposal.
• Skip free-space queries entirely for known-problematic backends (mtp://, gphoto2://).

File summaries

File	Description
src/nemo-places-sidebar.c	Adds async free-space querying with cancellation and unsafe-backend exclusion to avoid UI hangs.

Review details

Suppressed comments (1)

src/nemo-places-sidebar.c:534

• query_disk_full_async() stores a raw sidebar pointer in DiskFullQueryData. If the sidebar gets disposed while an async query is still in flight and a backend ignores cancellation (or completes after dispose), the callback can dereference freed memory. Take a strong reference when enqueuing the async request.

    data = g_new0 (DiskFullQueryData, 1);
    data->sidebar = sidebar;
    data->base_tooltip = g_strdup (base_tooltip);
    data->show_df_when_available = show_df_when_available;

• Files reviewed: 1/1 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🟡 Changes recommended

The async query user_data does not take a reference to sidebar despite unref’ing it later, which can lead to refcount underflow and use-after-free during/after dispose.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 1
• Review effort level: Lite

[Copilot]

🟡 Changes recommended

The new async query user-data holds a strong sidebar reference that can prevent teardown if a backend never completes, and the new tooltip assembly should escape strings since tooltips are rendered as markup.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 2
• Review effort level: Lite

[mtwebster]

Is show_df_when_available even necessary? It's only ever TRUE if df_file != NULL isn't it? So we wouldn't be running query_disk_full_async() on a location we didn't want show it in the first place...

[leigh123linux]

Home is the one place they diverge. df_file is always non-NULL for Home (we always want to query its free space, for the tooltip), but show_df_when_available there is home_on_different_fs(mount_uri) — independently FALSE when Home is on the same filesystem as root:

[leigh123linux]