Created attachment 174888 [details] [review]
iconcache: Ensure we don't lose data on power loss

Review of attachment 174888 [details] [review]:

Other than that, sounds ok.

::: gtk/updateiconcache.c
@@ +1497,3 @@
+      g_printerr (_("Failed to write cache file: %s\n"), g_strerror (errno));
+      exit (1);
+    }

If we want to be anal here, we probably don't want to ignore the return value of fclose either...

(In reply to comment #2)
> Review of attachment 174888 [details] [review]:
> 
> Other than that, sounds ok.
> 
> ::: gtk/updateiconcache.c
> @@ +1497,3 @@
> +      g_printerr (_("Failed to write cache file: %s\n"), g_strerror (errno));
> +      exit (1);
> +    }
> 
> If we want to be anal here, we probably don't want to ignore the return value
> of fclose either...

You're absolutely right, new patch incoming.

Created attachment 174891 [details] [review]
iconcache: Ensure we don't lose data on power loss

fsync() should ensure our data hits disk; since corrupt icon
caches break all apps, we need to ensure it's valid.

Review of attachment 174891 [details] [review]:

Looks good now.

Committed to gtk-2-24.  Note this patch could go into gtk3, but better to remove gtk-update-icon-cache-3.0 entirely.

Committed to master as commit 6c6b49392629a8ee2facafb66c8867a49a3e9036

(In reply to comment #6)
> Committed to gtk-2-24.  Note this patch could go into gtk3, but better to
> remove gtk-update-icon-cache-3.0 entirely.

Why?