https://bugzilla.gnome.org/show_bug.cgi?id=756431

There's a bug in the code where we ignore all renewal requests for
objects we already know about.

This means we'll only ever honor renewal requests that happen almost
immediately after start up.

This commit fixes the bug, so the only renewal requests that aren't
honored, are those that have been specifically disabled by the user.
This was clearly the original intention of the buggy code.

https://bugzilla.gnome.org/show_bug.cgi?id=756495

https://bugzilla.gnome.org/show_bug.cgi?id=756431

https://bugzilla.gnome.org/show_bug.cgi?id=768845

https://bugzilla.gnome.org/show_bug.cgi?id=768845

https://bugzilla.gnome.org/show_bug.cgi?id=762155

In the subsequent commit, we are going to use show_account_items in
goa_provider_ensure_credentials_sync. Let's use a suitable name to
clarify that it's no longer specific to goa_provider_show_account_real.

https://bugzilla.gnome.org/show_bug.cgi?id=762155

This isn't a problem at the moment, but every other implementation of
this virtual method does the same, so let's be consistent.

https://bugzilla.gnome.org/show_bug.cgi?id=762155

https://bugzilla.gnome.org/show_bug.cgi?id=762155

The gdbus-codegen-ed code doesn't use G_PARAM_EXPLICIT_NOTIFY. This
causes spurious notifications which can lead to the key file being
needlessly updated. Since The D-Bus objects are backed by the key
file, and are updated every time the key file changes, this can cause
an indefinite loop. Let's guard against that.

https://bugzilla.gnome.org/show_bug.cgi?id=765994

We can now use g_key_file_save_to_file which wraps g_key_file_to_data
and g_file_set_contents.

At the moment, g_key_file_to_data is documented as never throwing an
error. So we don't lose anything in terms of error reporting. Even if
this changes in the future, I think we can live without a separate
error message for g_key_file_to_data.

Bump minimum GLib version to 2.40.

https://bugzilla.gnome.org/show_bug.cgi?id=688041

We should identify temporary accounts as such by setting the
org.gnome.OnlineAccounts.Account:IsTemporary D-Bus property.

https://bugzilla.gnome.org/show_bug.cgi?id=765994

Bad things will happen if we set the GError to indicate failure, but
don't return FALSE.

We are setting the ret variable on each intermediate step, and if we
failed to look up smtp-password, it will be set to TRUE, which we will
return even if the GError is set. Let's only update the ret variable
when everything is done.

https://bugzilla.gnome.org/show_bug.cgi?id=760991

While at it, fix the same leak in refresh_account()

==9457== 11 bytes in 1 blocks are definitely lost in loss record 972 of 23,129
==9457==    at 0x4C28BF6: malloc (vg_replace_malloc.c:299)
==9457==    by 0x1456711C: g_malloc (gmem.c:94)
==9457==    by 0x145673FE: g_malloc_n (gmem.c:331)
==9457==    by 0x14581DCE: g_strdup (gstrfuncs.c:363)
==9457==    by 0x5BE4A17: goa_utils_parse_email_address (goautils.c:557)
==9457==    by 0x5BD62A5: add_account (goaimapsmtpprovider.c:1084)
==9457==    by 0x5BBD6A1: goa_provider_add_account (goaprovider.c:382)
==9457==    by 0x49C7D1: add_account_dialog_add_account (cc-online-accounts-add-account-dialog.c:83)
==9457==    by 0x49C864: list_box_row_activated_cb (cc-online-accounts-add-account-dialog.c:102)
==9457==    by 0x142CC6D4: g_cclosure_marshal_VOID__OBJECTv (gmarshal.c:2102)
==9457==    by 0x142C7C43: _g_closure_invoke_va (gclosure.c:867)
==9457==    by 0x142E2524: g_signal_emit_valist (gsignal.c:3294)
==9457==    by 0x142E369C: g_signal_emit (gsignal.c:3441)
==9457==    by 0x1264EAF2: gtk_list_box_activate (gtklistbox.c:1754)
==9457==    by 0x1264EB4B: gtk_list_box_select_and_activate_full (gtklistbox.c:1768)
==9457==    by 0x1264F115: gtk_list_box_multipress_gesture_released (gtklistbox.c:1964)
==9457==    by 0x1D52AD2F: ffi_call_unix64 (unix64.S:76)
==9457==    by 0x1D52A79A: ffi_call (ffi64.c:525)

https://bugzilla.gnome.org/show_bug.cgi?id=764283