Created attachment 240432 [details] [review]
window: reset timeout source id before setting property

gtk_window_set_mnemonics_visible() will try to g_source_remove() it
otherwise, which seems harmless, but conceptually wrong.

Created attachment 240433 [details] [review]
window: rename functions dealing with mnemonic delayed display

Don't mention "auto mnemonics", since those methods are purely about
scheduling a delayed display, and that makes understanding the code a
bit harder.

Created attachment 240434 [details] [review]
window: refactor a function

Instead of having maybe_set_mnemonics_visible(), separate the checks
from the actual scheduling of mnemonics display.

Created attachment 240435 [details] [review]
window: don't schedule a display of mnemonics if window has no modifier

When the window has no mnemonics modifier set, as in the case of a
GtkMenu, never schedule a display of mnemonics on focus-in.

Previously, for those windows, the GdkModifierType mask fetched from the
device would typically have been zero, leading to the

mnemonic_modifier == (mask & gtk_accelerator_get_default_mod_mask ())

check to succeed, so we would always trigger a display for popup menus.

Review of attachment 240432 [details] [review]:

sure

Review of attachment 240433 [details] [review]:

fine with me

Review of attachment 240434 [details] [review]:

ok

Review of attachment 240435 [details] [review]:

good catch

Attachment 240432 [details] pushed as a101e85 - window: reset timeout source id before setting property
Attachment 240433 [details] pushed as 1b3f5d3 - window: rename functions dealing with mnemonic delayed display
Attachment 240434 [details] pushed as 4a8203a - window: refactor a function
Attachment 240435 [details] pushed as 66b0b28 - window: don't schedule a display of mnemonics if window has no modifier